1 2020-03-04 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/94001
4 * tree-tailcall.c (process_assignment): Before comparing op1 to
5 *ass_var, verify *ass_var is non-NULL.
7 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
10 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
13 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
15 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
16 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
17 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
18 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
19 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
20 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
21 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
22 (V_bf_low, V_bf_cvt_m): New mode attributes.
23 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
24 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
25 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
26 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
27 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
29 2020-03-03 Jakub Jelinek <jakub@redhat.com>
31 PR tree-optimization/93582
32 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
33 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
34 members, initialize them in the constructor and if mask is non-NULL,
35 artificially push_partial_def {} for the portions of the mask that
37 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
38 val and return (void *)-1. Formatting fix.
39 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
41 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
42 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
44 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
46 (visit_stmt): Formatting fix.
48 2020-03-03 Richard Biener <rguenther@suse.de>
50 PR tree-optimization/93946
51 * alias.h (refs_same_for_tbaa_p): Declare.
52 * alias.c (refs_same_for_tbaa_p): New function.
53 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
55 * tree-ssa-scopedtables.h
56 (avail_exprs_stack::lookup_avail_expr): Add output argument
57 giving access to the hashtable entry.
58 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
60 * tree-ssa-dom.c: Include alias.h.
61 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
62 removing redundant store.
63 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
64 (ao_ref_init_from_vn_reference): Adjust prototype.
65 (vn_reference_lookup_pieces): Likewise.
66 (vn_reference_insert_pieces): Likewise.
67 * tree-ssa-sccvn.c: Track base alias set in addition to alias
69 (eliminate_dom_walker::eliminate_stmt): Also check base alias
70 set when removing redundant stores.
71 (visit_reference_op_store): Likewise.
72 * dse.c (record_store): Adjust valdity check for redundant
75 2020-03-03 Jakub Jelinek <jakub@redhat.com>
78 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
80 PR rtl-optimization/94002
81 * explow.c (plus_constant): Punt if cst has VOIDmode and
82 get_pool_mode is different from mode.
84 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
86 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
87 address has an offset which fits the scalling constraint for a
89 (legitimate_scaled_address_p): Update use
90 leigitimate_small_data_address_p.
91 (arc_print_operand): Likewise.
92 (arc_legitimate_address_p): Likewise.
93 (legitimate_small_data_address_p): Likewise.
95 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
97 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
98 (fnmasf4_fpu): Likewise.
100 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
102 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
105 (adddi3_i): Remove pattern.
106 (subdi3_i): Likewise.
108 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
110 * config/arc/arc.md (eh_return): Add length info.
112 2020-03-02 David Malcolm <dmalcolm@redhat.com>
114 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
116 2020-03-02 David Malcolm <dmalcolm@redhat.com>
118 * doc/invoke.texi (Static Analyzer Options): Add
119 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
122 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
125 * config/i386/i386.md (movstrict<mode>): Allow only
126 registers with VALID_INT_MODE_P modes.
128 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
130 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
131 (reduc_insn): Use 'U' and 'B' operand codes.
132 (reduc_<reduc_op>_scal_<mode>): Allow all types.
133 (reduc_<reduc_op>_scal_v64di): Delete.
134 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
135 (*plus_carry_dpp_shr_v64si): Change to ...
136 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
137 (mov_from_lane63_v64di): Change to ...
138 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
139 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
140 Support UNSPEC_MOV_DPP_SHR output formats.
141 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
142 Add "use_extends" reductions.
143 (print_operand_address): Add 'I' and 'U' codes.
144 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
146 2020-03-02 Martin Liska <mliska@suse.cz>
148 * lto-wrapper.c: Fix typo in comment about
149 C++ standard version.
151 2020-03-01 Martin Sebor <msebor@redhat.com>
154 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
156 2020-03-01 Martin Sebor <msebor@redhat.com>
159 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
160 of a pointer in the outermost ADDR_EXPRs.
162 2020-02-28 Jeff Law <law@redhat.com>
164 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
165 * config/v850/v850.c (v850_asm_trampoline_template): Update
168 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
171 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
174 2020-02-28 Martin Liska <mliska@suse.cz>
177 * configure.ac: Improve detection of ld_date by requiring
178 either two dashes or none.
179 * configure: Regenerate.
181 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
183 PR rtl-optimization/93564
184 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
185 do not honor reg alloc order.
187 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
190 * config/aarch64/aarch64.c (aarch64_override_options): Fix
191 misleading warning string.
193 2020-02-27 Martin Sebor <msebor@redhat.com>
195 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
197 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
200 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
201 Split the insn into two parts. This insn only does variable
202 extract from a register.
203 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
204 variable extract from memory.
205 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
206 only does variable extract from a register.
207 (vsx_extract_v4sf_var_load): New insn, do variable extract from
209 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
210 into two parts. This insn only does variable extract from a
212 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
213 do variable extract from memory.
215 2020-02-27 Martin Jambor <mjambor@suse.cz>
216 Feng Xue <fxue@os.amperecomputing.com>
219 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
220 new function calls_same_node_or_its_all_contexts_clone_p.
221 (cgraph_edge_brings_value_p): Use it.
222 (cgraph_edge_brings_value_p): Likewise.
223 (self_recursive_pass_through_p): Return false if caller is a clone.
224 (self_recursive_agg_pass_through_p): Likewise.
226 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
229 * alias.c (ends_tbaa_access_path_p): Break out from ...
230 (component_uses_parent_alias_set_from): ... here.
231 * alias.h (ends_tbaa_access_path_p): Declare.
232 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
233 handle trailing arrays past end of tbaa access path.
234 (aliasing_component_refs_p): ... here; likewise.
235 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
236 path; disambiguate also past end of it.
237 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
240 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
242 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
243 beginning of the file.
244 (vcreate_bf16, vcombine_bf16): New.
245 (vdup_n_bf16, vdupq_n_bf16): New.
246 (vdup_lane_bf16, vdup_laneq_bf16): New.
247 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
248 (vduph_lane_bf16, vduph_laneq_bf16): New.
249 (vset_lane_bf16, vsetq_lane_bf16): New.
250 (vget_lane_bf16, vgetq_lane_bf16): New.
251 (vget_high_bf16, vget_low_bf16): New.
252 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
253 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
254 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
255 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
256 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
257 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
258 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
259 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
260 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
261 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
262 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
263 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
264 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
265 (vreinterpretq_bf16_p128): New.
266 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
267 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
268 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
269 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
270 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
271 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
272 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
273 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
274 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
275 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
276 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
277 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
278 (vreinterpretq_p128_bf16): New.
279 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
281 (V_elem_l): Likewise.
284 (V_DOUBLE): Likewise.
285 (VDQX): Add V4BF and V8BF.
286 (V_two_elem, V_three_elem, V_four_elem): Likewise.
289 (V_double_vector_mode): Likewise.
290 (V_cmp_result): Likewise.
291 (V_uf_sclr): Likewise.
292 (V_sz_elem): Likewise.
293 (Is_d_reg): Likewise.
294 (V_mode_nunits): Likewise.
295 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
297 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
299 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
300 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
301 (<expander><mode>3<exec>): Likewise.
302 (<expander><mode>3): New.
303 (v<expander><mode>3): New.
304 (<expander><mode>3): New.
305 (<expander><mode>3<exec>): Rename to ...
306 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
307 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
309 2020-02-27 Alexandre Oliva <oliva@adacore.com>
311 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
314 2020-02-27 Richard Biener <rguenther@suse.de>
316 PR tree-optimization/93508
317 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
318 non-_CHK variants. Valueize their length arguments.
320 2020-02-27 Richard Biener <rguenther@suse.de>
322 PR tree-optimization/93953
323 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
324 to the hash-map entry.
326 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
328 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
330 2020-02-27 Mark Williams <mwilliams@fb.com>
332 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
333 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
334 -ffile-prefix-map and -fmacro-prefix-map.
335 * lto-streamer-out.c: Include file-prefix-map.h.
336 (lto_output_location): Remap the file part of locations.
338 2020-02-27 Jakub Jelinek <jakub@redhat.com>
341 * gimplify.c (gimplify_init_constructor): Don't promote readonly
342 DECL_REGISTER variables to TREE_STATIC.
344 PR tree-optimization/93582
345 PR tree-optimization/93945
346 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
347 non-zero INTEGER_CST second argument and ref->offset or ref->size
348 not a multiple of BITS_PER_UNIT.
350 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
352 * doc/install.texi (Binaries): Update description of BullFreeware.
354 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
358 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
359 C++ Language Options, Warning Options, and Static Analyzer
360 Options lists. Document negative form of options enabled by
361 default. Move some things around to more accurately sort
362 warnings by category.
363 (C++ Dialect Options, Warning Options, Static Analyzer
364 Options): Document negative form of options when enabled by
365 default. Move some things around to more accurately sort
366 warnings by category. Add some missing index entries.
369 2020-02-26 Carl Love <cel@us.ibm.com>
372 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
373 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
374 for the vector unsigned short arguments. It is also listed as the
375 name of the built-in for arguments vector unsigned short,
376 vector unsigned int and vector unsigned long long built-ins. The
377 name of the builtins for these arguments should be:
378 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
379 __builtin_crypto_vpmsumd respectively.
381 2020-02-26 Richard Biener <rguenther@suse.de>
383 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
384 and load permutation.
386 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
389 * optabs-tree.c (supportable_convert_operation): Reject types with
392 2020-02-26 David Malcolm <dmalcolm@redhat.com>
394 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
396 2020-02-26 Jakub Jelinek <jakub@redhat.com>
398 PR tree-optimization/93820
399 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
400 argument to ALL_INTEGER_CST_P boolean.
401 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
402 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
403 adjacent INTEGER_CST store into merged_store->only_constants like
406 2020-02-25 Jakub Jelinek <jakub@redhat.com>
409 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
411 * cfghooks.c (verify_flow_info): Likewise.
412 * predict.c (combine_predictions_for_bb): Likewise.
413 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
414 sucessor -> successor.
415 (find_traces_1_round): Fix comment typo, destinarion -> destination.
416 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
418 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
419 message typo, sucessors -> successors.
421 2020-02-25 Martin Sebor <msebor@redhat.com>
423 * doc/extend.texi (attribute access): Correct an example.
425 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
427 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
429 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
431 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
432 (VD): Enable for V4BF.
434 (VQ): Enable for V8BF.
437 (VDBL, Vdbl): Add V4BF.
438 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
439 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
440 (bfloat16x8x2_t): Likewise.
441 (bfloat16x4x3_t): Likewise.
442 (bfloat16x8x3_t): Likewise.
443 (bfloat16x4x4_t): Likewise.
444 (bfloat16x8x4_t): Likewise.
445 (vcombine_bf16): New.
446 (vld1_bf16, vld1_bf16_x2): New.
447 (vld1_bf16_x3, vld1_bf16_x4): New.
448 (vld1q_bf16, vld1q_bf16_x2): New.
449 (vld1q_bf16_x3, vld1q_bf16_x4): New.
450 (vld1_lane_bf16): New.
451 (vld1q_lane_bf16): New.
452 (vld1_dup_bf16): New.
453 (vld1q_dup_bf16): New.
456 (vld2_dup_bf16): New.
457 (vld2q_dup_bf16): New.
460 (vld3_dup_bf16): New.
461 (vld3q_dup_bf16): New.
464 (vld4_dup_bf16): New.
465 (vld4q_dup_bf16): New.
466 (vst1_bf16, vst1_bf16_x2): New.
467 (vst1_bf16_x3, vst1_bf16_x4): New.
468 (vst1q_bf16, vst1q_bf16_x2): New.
469 (vst1q_bf16_x3, vst1q_bf16_x4): New.
470 (vst1_lane_bf16): New.
471 (vst1q_lane_bf16): New.
479 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
481 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
482 (VALL_F16): Likewise.
483 (VALLDI_F16): Likewise.
486 (vswap_width_name): Likewise.
487 (VSWAP_WIDTH): Likewise.
491 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
492 (vget_lane_bf16, vgetq_lane_bf16): New.
494 (vdup_n_bf16, vdupq_n_bf16): New.
495 (vdup_lane_bf16, vdup_laneq_bf16): New.
496 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
497 (vduph_lane_bf16, vduph_laneq_bf16): New.
498 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
499 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
500 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
501 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
502 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
503 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
504 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
505 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
506 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
507 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
508 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
509 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
510 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
511 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
512 (vreinterpretq_bf16_p128): New.
513 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
514 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
515 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
516 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
517 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
518 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
519 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
520 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
521 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
522 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
523 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
524 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
525 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
526 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
527 (vreinterpretq_p128_bf16): New.
529 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
531 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
532 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
533 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
534 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
535 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
536 * config/arm/iterators.md (VSF2BF): New attribute.
537 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
538 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
539 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
541 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
543 * config/arm/arm.md (required_for_purecode): New attribute.
544 (enabled): Handle required_for_purecode.
545 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
546 work with -mpure-code.
548 2020-02-25 Jakub Jelinek <jakub@redhat.com>
550 PR rtl-optimization/93908
551 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
554 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
556 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
558 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
560 * doc/install.texi (--enable-checking): Adjust wording.
562 2020-02-25 Richard Biener <rguenther@suse.de>
564 PR tree-optimization/93868
565 * tree-vect-slp.c (slp_copy_subtree): New function.
566 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
567 re-arranging stmts in it.
569 2020-02-25 Jakub Jelinek <jakub@redhat.com>
572 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
573 dummy function and remove it at the end.
576 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
577 paramter -> parameter.
578 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
579 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
581 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
583 * doc/install.texi (--enable-checking): Properly document current
585 (--enable-stage1-checking): Minor clarification about bootstrap.
587 2020-02-24 David Malcolm <dmalcolm@redhat.com>
590 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
591 -fanalyzer-checker=taint is also required.
592 (-fanalyzer-checker=): Note that providing this option enables the
593 given checker, and doing so may be required for checkers that are
596 2020-02-24 David Malcolm <dmalcolm@redhat.com>
598 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
599 significant control flow events; add a "3" which shows all
600 control flow events; the old "3" becomes "4".
602 2020-02-24 Jakub Jelinek <jakub@redhat.com>
604 PR tree-optimization/93582
605 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
606 pd.offset and pd.size to be counted in bits rather than bytes, add
607 support for maxsizei that is not a multiple of BITS_PER_UNIT and
608 handle bitfield stores and loads.
609 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
610 uncomparable quantities - bytes vs. bits. Allow push_partial_def
611 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
612 pd.offset/pd.size to be counted in bits rather than bytes.
613 Formatting fix. Rename shadowed len variable to buflen.
615 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
616 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
619 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
620 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
621 * opts-common.c (parse_options_from_collect_gcc_options): New function.
622 (prepend_xassembler_to_collect_as_options): Likewise.
623 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
624 (prepend_xassembler_to_collect_as_options): Likewise.
625 * lto-opts.c (lto_write_options): Stream assembler options
626 in COLLECT_AS_OPTIONS.
627 * lto-wrapper.c (xassembler_options_error): New static variable.
628 (get_options_from_collect_gcc_options): Move parsing options code to
629 parse_options_from_collect_gcc_options and call it.
630 (merge_and_complain): Validate -Xassembler options.
631 (append_compiler_options): Handle OPT_Xassembler.
632 (run_gcc): Append command line -Xassembler options to
634 * doc/invoke.texi: Add documentation about using Xassembler
637 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
639 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
641 (riscv_rtx_costs): Update cost model for LTGT.
643 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
645 PR rtl-optimization/93564
646 * ira-color.c (struct update_cost_queue_elem): New member start.
647 (queue_update_cost, get_next_update_cost): Add new arg start.
648 (allocnos_conflict_p): New function.
649 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
650 Add checking conflicts with allocnos_conflict_p.
651 (update_costs_from_prefs, restore_costs_from_copies): Adjust
652 update_costs_from_allocno calls.
653 (update_conflict_hard_regno_costs): Add checking conflicts with
654 allocnos_conflict_p. Adjust calls of queue_update_cost and
655 get_next_update_cost.
656 (assign_hard_reg): Adjust calls of queue_update_cost. Add
658 (bucket_allocno_compare_func): Restore previous version.
660 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
662 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
663 double-word size when handling aggregate return values.
664 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
665 that homogeneous SFmode and DFmode aggregates are passed and returned
666 in general registers.
668 2020-02-21 Jakub Jelinek <jakub@redhat.com>
671 * opts.c (print_filtered_help): Translate help before appending
672 messages to it rather than after that.
674 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
676 PR rtl-optimization/PR92989
677 * lra-lives.c (process_bb_lives): Restore the original order
678 of the bb liveness update. Call make_hard_regno_dead for each
679 register clobbered at the start of an EH receiver.
681 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
684 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
685 self-recursively generated.
687 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
690 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
693 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
695 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
696 Document new target supports option.
698 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
700 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
701 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
702 * config/arm/iterators.md (MATMUL): New iterator.
703 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
704 (mmla_sfx): New attribute.
705 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
706 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
707 (UNSPEC_MATMUL_US): New.
709 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
711 * config/arm/arm.md: Prevent scalar shifts from being used when big
714 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
715 Richard Biener <rguenther@suse.de>
717 PR tree-optimization/93586
718 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
719 after mismatched array refs; do not sure type size information to
720 recover from unmatched referneces with !flag_strict_aliasing_p.
722 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
724 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
725 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
726 (scatter_store<mode>): Rename to ...
727 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
728 (scatter<mode>_exec): Delete. Move contents ...
729 (mask_scatter_store<mode>): ... here, and rename that to ...
730 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
731 Remove mode conversion.
732 (mask_gather_load<mode>): Rename to ...
733 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
734 Remove mode conversion.
735 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
737 2020-02-21 Martin Jambor <mjambor@suse.cz>
739 PR tree-optimization/93845
740 * tree-sra.c (verify_sra_access_forest): Only test access size of
743 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
745 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
746 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
747 (addv64di3_exec): Likewise.
748 (subv64di3): Likewise.
749 (subv64di3_exec): Likewise.
750 (addv64di3_zext): Likewise.
751 (addv64di3_zext_exec): Likewise.
752 (addv64di3_zext_dup): Likewise.
753 (addv64di3_zext_dup_exec): Likewise.
754 (addv64di3_zext_dup2): Likewise.
755 (addv64di3_zext_dup2_exec): Likewise.
756 (addv64di3_sext_dup2): Likewise.
757 (addv64di3_sext_dup2_exec): Likewise.
758 (<expander>v64di3): Likewise.
759 (<expander>v64di3_exec): Likewise.
760 (*<reduc_op>_dpp_shr_v64di): Likewise.
761 (*plus_carry_dpp_shr_v64di): Likewise.
762 * config/gcn/gcn.md (adddi3): Likewise.
763 (addptrdi3): Likewise.
764 (<expander>di3): Likewise.
766 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
768 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
770 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
772 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
773 support. Use aarch64_emit_mult instead of emitting multiplication
774 instructions directly.
775 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
776 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
778 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
780 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
781 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
782 instead of emitting multiplication instructions directly.
783 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
784 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
785 (@aarch64_frecps<mode>): New expanders.
787 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
789 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
790 on and produce uint64_ts rather than ints.
791 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
792 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
794 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
796 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
797 an unused xmsk register when handling approximate rsqrt.
799 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
801 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
802 flag_finite_math_only condition.
804 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
807 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
808 to destination operand for shufps alternative.
809 (*vec_extractv2si_1): Ditto.
811 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
814 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
817 2020-02-20 Martin Liska <mliska@suse.cz>
820 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
822 2020-02-20 Martin Liska <mliska@suse.cz>
825 * common/config/avr/avr-common.c: Remote trailing "|".
827 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
829 * collect2.c (maybe_run_lto_and_relink): Fix typo in
832 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
834 PR tree-optimization/93767
835 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
836 access-size bias from the offset calculations for negative strides.
838 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
840 * collect2.c (c_file, o_file): Make const again.
841 (ldout,lderrout, dump_ld_file): Remove.
842 (tool_cleanup): Avoid calling not signal-safe functions.
843 (maybe_run_lto_and_relink): Avoid possible signal handler
844 access to unintialzed memory (lto_o_files).
845 (main): Avoid leaking temp files in $TMPDIR.
846 Initialize c_file/o_file with concat, which avoids exposing
847 uninitialized memory to signal handler, which calls unlink(!).
848 Avoid calling maybe_unlink when the main function returns,
849 since the atexit handler is already doing this.
850 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
852 2020-02-19 Martin Jambor <mjambor@suse.cz>
854 PR tree-optimization/93776
855 * tree-sra.c (create_access): Do not create zero size accesses.
856 (get_access_for_expr): Do not search for zero sized accesses.
858 2020-02-19 Martin Jambor <mjambor@suse.cz>
860 PR tree-optimization/93667
861 * tree-sra.c (scalarizable_type_p): Return false if record fields
862 do not follow wach other.
864 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
866 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
867 rather than fmv.x.s/fmv.s.x.
869 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
871 * config/aarch64/aarch64-simd-builtins.def
872 (intrinsic_vec_smult_lo_): New.
873 (intrinsic_vec_umult_lo_): Likewise.
874 (vec_widen_smult_hi_): Likewise.
875 (vec_widen_umult_hi_): Likewise.
876 * config/aarch64/aarch64-simd.md
877 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
878 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
879 (vmull_high_s16): Likewise.
880 (vmull_high_s32): Likewise.
881 (vmull_high_u8): Likewise.
882 (vmull_high_u16): Likewise.
883 (vmull_high_u32): Likewise.
884 (vmull_s8): Likewise.
885 (vmull_s16): Likewise.
886 (vmull_s32): Likewise.
887 (vmull_u8): Likewise.
888 (vmull_u16): Likewise.
889 (vmull_u32): Likewise.
891 2020-02-18 Martin Liska <mliska@suse.cz>
893 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
894 bootstrap by missing removal of invalid sanity check.
896 2020-02-18 Martin Liska <mliska@suse.cz>
899 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
900 Always compare LHS of gimple_assign.
902 2020-02-18 Martin Liska <mliska@suse.cz>
905 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
906 and return type of functions.
907 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
908 Drop MALLOC attribute for void functions.
909 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
910 malloc_state for a new VOID clone.
912 2020-02-18 Martin Liska <mliska@suse.cz>
915 * common.opt: Add -fprofile-reproducibility.
916 * doc/invoke.texi: Document it.
917 * value-prof.c (dump_histogram_value):
918 Document and support behavior for counters[0]
919 being a negative value.
920 (get_nth_most_common_value): Handle negative
921 counters[0] in respect to flag_profile_reproducible.
923 2020-02-18 Jakub Jelinek <jakub@redhat.com>
926 * cgraph.c (verify_speculative_call): Use speculative_id instead of
927 speculative_uid in messages. Remove trailing whitespace from error
928 message. Use num_speculative_call_targets instead of
929 num_speculative_targets in a message.
930 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
931 edge messages and stmt instead of cal_stmt in reference message.
933 PR tree-optimization/93780
934 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
935 before calling build_vector_type.
936 (execute_update_addresses_taken): Likewise.
939 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
940 typo, functoin -> function.
941 * tree.c (free_lang_data_in_decl): Fix comment typo,
942 functoin -> function.
943 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
945 2020-02-17 David Malcolm <dmalcolm@redhat.com>
947 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
949 (print_option_information): Don't call get_option_url if URLs
952 2020-02-17 Alexandre Oliva <oliva@adacore.com>
954 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
955 handling of register_common-less targets.
957 2020-02-17 Martin Liska <mliska@suse.cz>
960 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
962 2020-02-17 Martin Liska <mliska@suse.cz>
965 * config/rs6000/rs6000.c (rs6000_option_override_internal):
968 2020-02-17 Martin Liska <mliska@suse.cz>
971 * config/rx/elf.opt: Fix typo.
973 2020-02-17 Richard Biener <rguenther@suse.de>
976 * opts-global.c (print_ignored_options): Use inform and
979 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
982 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
984 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
987 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
988 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
990 2020-02-15 Jason Merrill <jason@redhat.com>
992 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
994 2020-02-15 Jakub Jelinek <jakub@redhat.com>
996 PR tree-optimization/93744
997 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
998 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
999 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1000 sure @2 in the first and @1 in the other patterns has no side-effects.
1002 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1003 Bernd Edlinger <bernd.edlinger@hotmail.de>
1007 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1008 * configure.ac (--with-diagnostics-urls): New configuration
1009 option, based on --with-diagnostics-color.
1010 (DIAGNOSTICS_URLS_DEFAULT): New define.
1011 * config.h: Regenerate.
1012 * configure: Regenerate.
1013 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1014 DIAGNOSTICS_URLS_DEFAULT from configure-time
1015 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1016 and TERM_URLS environment variable.
1017 * diagnostic-url.h (diagnostic_url_format): New enum type.
1018 (diagnostic_urls_enabled_p): rename to...
1019 (determine_url_format): ... this, and change return type.
1020 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1021 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1022 the linux console, and mingw.
1023 (diagnostic_urls_enabled_p): rename to...
1024 (determine_url_format): ... this, and adjust.
1025 * pretty-print.h (pretty_printer::show_urls): rename to...
1026 (pretty_printer::url_format): ... this, and change to enum.
1027 * pretty-print.c (pretty_printer::pretty_printer,
1028 pp_begin_url, pp_end_url, test_urls): Adjust.
1029 * doc/install.texi (--with-diagnostics-urls): Document the new
1030 configuration option.
1031 (--with-diagnostics-color): Document the existing interaction
1032 with GCC_COLORS better.
1033 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1034 vindex reference. Update description of defaults based on the above.
1035 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1036 interacts with GCC_COLORS.
1038 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1041 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1042 conjunction with TARGET_GNU_TLS in early return.
1044 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1046 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1047 the mode is not wider than UNITS_PER_WORD.
1049 2020-02-14 Martin Jambor <mjambor@suse.cz>
1051 PR tree-optimization/93516
1052 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1053 access of the same type as the parent.
1054 (propagate_subaccesses_from_lhs): Likewise.
1056 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1059 * config/i386/avx512vbmi2intrin.h
1060 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1061 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1062 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1063 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1064 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1065 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1066 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1067 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1068 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1069 of lacking a closing parenthesis.
1070 * config/i386/avx512vbmi2vlintrin.h
1071 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1072 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1073 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1074 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1075 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1076 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1077 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1078 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1079 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1080 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1081 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1082 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1083 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1084 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1085 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1086 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1087 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1088 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1090 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1093 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1094 the target function entry.
1096 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1098 * common/config/arc/arc-common.c (arc_option_optimization_table):
1099 Disable if-conversion step when optimized for size.
1101 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1103 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1104 R12-R15 are always in ARCOMPACT16_REGS register class.
1105 * config/arc/arc.opt (mq-class): Deprecate.
1106 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1108 * doc/invoke.texi (mq-class): Update text.
1109 * common/config/arc/arc-common.c (arc_option_optimization_table):
1112 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1114 * config/arc/arc.c (arc_insn_cost): New function.
1115 (TARGET_INSN_COST): Define.
1116 * config/arc/arc.md (cost): New attribute.
1117 (add_n): Use arc_nonmemory_operand.
1118 (ashlsi3_insn): Likewise, also update constraints.
1119 (ashrsi3_insn): Likewise.
1120 (rotrsi3): Likewise.
1121 (add_shift): Likewise.
1122 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1124 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1126 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1128 (umulsidi_600): Likewise.
1130 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1133 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1134 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1135 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1136 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1137 pass __A to the builtin followed by __W instead of __A followed by
1139 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1140 _mm512_mask_popcnt_epi64): Likewise.
1141 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1142 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1143 _mm256_mask_popcnt_epi64): Likewise.
1145 PR tree-optimization/93582
1146 * fold-const.h (shift_bytes_in_array_left,
1147 shift_bytes_in_array_right): Declare.
1148 * fold-const.c (shift_bytes_in_array_left,
1149 shift_bytes_in_array_right): New function, moved from
1150 gimple-ssa-store-merging.c, no longer static.
1151 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1152 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1153 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1154 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1155 shift_bytes_in_array.
1156 (verify_shift_bytes_in_array): Rename to ...
1157 (verify_shift_bytes_in_array_left): ... this. Use
1158 shift_bytes_in_array_left instead of shift_bytes_in_array.
1159 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1160 instead of verify_shift_bytes_in_array.
1161 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1162 / native_interpret_expr where the store covers all needed bits,
1163 punt on PDP-endian, otherwise allow all involved offsets and sizes
1164 not to be byte-aligned.
1167 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1168 use const_0_to_255_operand predicate instead of immediate_operand.
1169 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1170 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1171 vgf2p8affineinvqb_<mode><mask_name>,
1172 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1173 const_0_to_255_operand predicated operands.
1175 2020-02-12 Jeff Law <law@redhat.com>
1177 * config/h8300/h8300.md (comparison shortening peepholes): Use
1178 a mode iterator to merge the HImode and SImode peepholes.
1180 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1183 * real.c (is_even): Make static. Function comment fix.
1184 (is_halfway_below): Make static, don't assert R is not inf/nan,
1185 instead return false for those. Small formatting fixes.
1187 2020-02-12 Martin Sebor <msebor@redhat.com>
1190 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1191 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1192 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1193 (strlen_check_and_optimize_call): Adjust callee name.
1195 2020-02-12 Jeff Law <law@redhat.com>
1197 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1198 (and (xor)) variant. Combine other two into single peephole.
1200 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1202 PR rtl-optimization/93565
1203 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1205 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1207 * config/aarch64/aarch64-simd.md
1208 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1209 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1210 generating separate ADDV and zero_extend patterns.
1211 * config/aarch64/iterators.md (VDQV_E): New iterator.
1213 2020-02-12 Jeff Law <law@redhat.com>
1215 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1216 expanders, splits, etc.
1217 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1218 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1219 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1220 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1221 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1222 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1224 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1226 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1229 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1230 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1231 TARGET_AVX512DQ from condition.
1232 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1233 instead of <mask_mode512bit_condition> in condition. If
1234 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1236 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1239 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1242 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1244 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1246 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1247 where strlen is more legible.
1248 (rs6000_builtin_vectorized_libmass): Ditto.
1249 (rs6000_print_options_internal): Ditto.
1251 2020-02-11 Martin Sebor <msebor@redhat.com>
1253 PR tree-optimization/93683
1254 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1256 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1258 * config/rs6000/predicates.md (cint34_operand): Rename the
1259 -mprefixed-addr option to be -mprefixed.
1260 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1261 the -mprefixed-addr option to be -mprefixed.
1262 (OTHER_FUTURE_MASKS): Likewise.
1263 (POWERPC_MASKS): Likewise.
1264 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1265 the -mprefixed-addr option to be -mprefixed. Change error
1266 messages to refer to -mprefixed.
1267 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1269 (rs6000_legitimate_offset_address_p): Likewise.
1270 (rs6000_mode_dependent_address): Likewise.
1271 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1272 "-mprefixed" for target attributes and pragmas.
1273 (address_to_insn_form): Rename the -mprefixed-addr option to be
1275 (rs6000_adjust_insn_length): Likewise.
1276 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1277 -mprefixed-addr option to be -mprefixed.
1278 (ASM_OUTPUT_OPCODE): Likewise.
1279 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1280 -mprefixed-addr option to be -mprefixed.
1281 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1282 -mprefixed-addr option to be prefixed. Change the option from
1283 being undocumented to being documented.
1284 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1285 -mprefixed option. Update the -mpcrel documentation to mention
1288 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1290 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1291 including FIRST_PSEUDO_REGISTER - 1.
1292 * ira-color.c (print_hard_reg_set): Ditto.
1294 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1296 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1297 (USTERNOP_QUALIFIERS): New define.
1298 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1299 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1300 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1301 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1302 * config/arm/arm_neon.h (vusdot_s32): New.
1303 (vusdot_lane_s32): New.
1304 (vusdotq_lane_s32): New.
1305 (vsudot_lane_s32): New.
1306 (vsudotq_lane_s32): New.
1307 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1308 * config/arm/iterators.md (DOTPROD_I8MM): New.
1309 (sup, opsuffix): Add <us/su>.
1310 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1311 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1313 2020-02-11 Richard Biener <rguenther@suse.de>
1315 PR tree-optimization/93661
1316 PR tree-optimization/93662
1317 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1319 * tree-sra.c (get_access_for_expr): Likewise.
1321 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1324 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1325 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1326 Change condition from TARGET_AVX2 to TARGET_AVX.
1328 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1331 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1332 argument of strncmp.
1334 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1336 Try to generate zero-based comparisons.
1337 * config/cris/cris.c (cris_reduce_compare): New function.
1338 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1339 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1340 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1342 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1345 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1346 in Thumb state and also as a destination in Arm state. Add T16
1349 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1351 * md.texi (Define Subst): Match closing paren in example.
1353 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1357 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1358 arguments of strncmp.
1360 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1363 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1364 but different source value.
1365 (adjust_callers_for_value_intersection): New function.
1366 (gather_edges_for_value): Adjust order of callers to let a
1367 non-self-recursive caller be the first element.
1368 (self_recursive_pass_through_p): Add a new parameter "simple", and
1369 check generalized self-recursive pass-through jump function.
1370 (self_recursive_agg_pass_through_p): Likewise.
1371 (find_more_scalar_values_for_callers_subset): Compute value from
1372 pass-through jump function for self-recursive.
1373 (intersect_with_plats): Cleanup previous implementation code for value
1374 itersection with self-recursive call edge.
1375 (intersect_with_agg_replacements): Likewise.
1376 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1377 function for self-recursive call edge. Cleanup previous implementation
1378 code for value intersection with self-recursive call edge.
1379 (decide_whether_version_node): Remove dead callers and adjust order
1380 to let a non-self-recursive caller be the first element.
1382 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1384 * recog.c: Move pass_split_before_sched2 code in front of
1385 pass_split_before_regstack.
1386 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1387 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1388 (rest_of_handle_split_before_sched2): Remove.
1389 (pass_split_before_sched2::execute): Unconditionally call
1391 (enable_split_before_sched2): New function.
1392 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1393 (pass_split_before_regstack::gate): Ditto.
1394 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1395 Update name check for renamed split4 pass.
1396 * config/sh/sh.c (register_sh_passes): Update pass insertion
1397 point for renamed split4 pass.
1399 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1401 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1402 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1403 copying them around between host and target.
1405 2020-02-08 Andrew Pinski <apinski@marvell.com>
1408 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1409 STRICT_ALIGNMENT also.
1411 2020-02-08 Jim Wilson <jimw@sifive.com>
1414 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1416 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1417 Jakub Jelinek <jakub@redhat.com>
1420 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1421 xmm16-xmm31 call-used even in 64-bit ms-abi.
1423 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1425 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1426 (simd_ummla, simd_usmmla): Likewise.
1427 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1428 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1429 (vusmmlaq_s32): New.
1431 2020-02-07 Richard Biener <rguenther@suse.de>
1434 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1435 skipping unreachable regions.
1436 (optimize_inline_calls): Skip folding stmts when we didn't
1439 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1442 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1443 Don't return aggregates with only SFmode and DFmode in SSE
1445 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1447 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1450 * config/rs6000/rs6000-logue.c
1451 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1452 if it fails, move rs into end_addr and retry. Add
1453 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1454 the insn pattern doesn't describe well what exactly happens to
1458 * config/i386/predicates.md (avx_identity_operand): Remove.
1459 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1460 (avx_<castmode><avxsizesuffix>_<castmode>,
1461 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1462 a VEC_CONCAT of the operand and UNSPEC_CAST.
1463 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1464 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1468 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1469 recog_data.insn if distance_non_agu_define changed it.
1471 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1474 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1475 we only had X-FORM (reg+reg) addressing for vectors. Also before
1476 ISA 3.0, we only had X-FORM addressing for scalars in the
1477 traditional Altivec registers.
1479 2020-02-06 <zhongyunde@huawei.com>
1480 Vladimir Makarov <vmakarov@redhat.com>
1482 PR rtl-optimization/93561
1483 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1484 hard register range.
1486 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1488 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1491 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1493 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1494 where the low and the high 32 bits are equal to each other specially,
1495 with an rldimi instruction.
1497 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1499 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1501 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1503 * config/arm/arm-tables.opt: Regenerate.
1505 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1508 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1509 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1510 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1512 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1514 PR rtl-optimization/87763
1515 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1517 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1519 * config/aarch64/aarch64-simd-builtins.def
1520 (bfmlaq): New built-in function.
1521 (bfmlalb): New built-in function.
1522 (bfmlalt): New built-in function.
1523 (bfmlalb_lane): New built-in function.
1524 (bfmlalt_lane): New built-in function.
1525 * config/aarch64/aarch64-simd.md
1526 (aarch64_bfmmlaqv4sf): New pattern.
1527 (aarch64_bfmlal<bt>v4sf): New pattern.
1528 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1529 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1530 (vbfmlalbq_f32): New intrinsic.
1531 (vbfmlaltq_f32): New intrinsic.
1532 (vbfmlalbq_lane_f32): New intrinsic.
1533 (vbfmlaltq_lane_f32): New intrinsic.
1534 (vbfmlalbq_laneq_f32): New intrinsic.
1535 (vbfmlaltq_laneq_f32): New intrinsic.
1536 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1537 (bt): New int attribute.
1539 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1541 * config/i386/i386.md (*pushtf): Emit "#" instead of
1542 calling gcc_unreachable in insn output.
1545 (*pushsf_rex64): Ditto for alternatives other than 1.
1546 (*pushsf): Ditto for alternatives other than 1.
1548 2020-02-06 Martin Liska <mliska@suse.cz>
1550 PR gcov-profile/91971
1551 PR gcov-profile/93466
1552 * coverage.c (coverage_init): Revert mangling of
1553 path into filename. It can lead to huge filename length.
1554 Creation of subfolders seem more natural.
1556 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1559 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1560 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1561 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1563 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1566 * config/i386/predicates.md (avx_identity_operand): New predicate.
1567 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1568 define_insn_and_split.
1571 * omp-low.c (use_pointer_for_field): For nested constructs, also
1572 look for map clauses on target construct.
1573 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1574 taskreg_nesting_level.
1577 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1578 shared clause, call omp_notice_variable on outer context if any.
1580 2020-02-05 Jason Merrill <jason@redhat.com>
1583 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1584 non-zero address even if weak and not yet defined.
1586 2020-02-05 Martin Sebor <msebor@redhat.com>
1588 PR tree-optimization/92765
1589 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1590 * tree-ssa-strlen.c (compute_string_length): Remove.
1591 (determine_min_objsize): Remove.
1592 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1593 Avoid using type size as the upper bound on string length.
1594 (handle_builtin_string_cmp): Add an argument. Adjust.
1595 (strlen_check_and_optimize_call): Pass additional argument to
1596 handle_builtin_string_cmp.
1598 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1600 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1601 (*pushdi2_rex64 peephole2): Unconditionally split after
1603 (*ashl<mode>3_doubleword): Ditto.
1604 (*<shift_insn><mode>3_doubleword): Ditto.
1606 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1609 * config/rs6000/rs6000.c (get_vector_offset): Fix
1611 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1613 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1615 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1618 (Special Functions for Debugging the Analyzer): Update description
1619 of __analyzer_dump_exploded_nodes.
1621 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1624 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1625 include sets and not clobbers in the vzeroupper pattern.
1626 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1627 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1628 (*avx_vzeroupper_1): New define_insn_and_split.
1631 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1632 don't run when !optimize.
1633 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1636 2020-02-05 Richard Biener <rguenther@suse.de>
1639 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1640 checks before matching calls.
1642 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1644 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1645 function comment typo.
1648 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1649 simd_clone_create failed when i == 0, adjust clone->nargs by
1652 2020-02-05 Martin Liska <mliska@suse.cz>
1655 * doc/invoke.texi: Document that one should
1656 not combine ASLR and -fpch.
1658 2020-02-04 Richard Biener <rguenther@suse.de>
1660 PR tree-optimization/93538
1661 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1663 2020-02-04 Richard Biener <rguenther@suse.de>
1665 PR tree-optimization/91123
1666 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1667 (vn_walk_cb_data::last_vuse): New member.
1668 (vn_walk_cb_data::saved_operands): Likewsie.
1669 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1670 (vn_walk_cb_data::push_partial_def): Use finish.
1671 (vn_reference_lookup_2): Update last_vuse and use finish if
1672 we've saved operands.
1673 (vn_reference_lookup_3): Use finish and update calls to
1674 push_partial_defs everywhere. When translating through
1675 memcpy or aggregate copies save off operands and alias-set.
1676 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1677 operation for redundant store removal.
1679 2020-02-04 Richard Biener <rguenther@suse.de>
1681 PR tree-optimization/92819
1682 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1683 generating more stmts than before.
1685 2020-02-04 Martin Liska <mliska@suse.cz>
1687 * config/arm/arm.c (arm_gen_far_branch): Move the function
1688 outside of selftests.
1690 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1692 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1693 function to adjust PC-relative vector addresses.
1694 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1695 handle vectors with PC-relative addresses.
1697 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1699 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1701 (hard_reg_and_mode_to_addr_mask): Delete.
1702 (rs6000_adjust_vec_address): If the original vector address
1703 was REG+REG or REG+OFFSET and the element is not zero, do the add
1704 of the elements in the original address before adding the offset
1705 for the vector element. Use address_to_insn_form to validate the
1706 address using the register being loaded, rather than guessing
1707 whether the address is a DS-FORM or DQ-FORM address.
1709 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1711 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1712 to calculate the offset in memory from the start of a vector of a
1713 particular element. Add code to keep the element number in
1714 bounds if the element number is variable.
1715 (rs6000_adjust_vec_address): Move calculation of offset of the
1716 vector element to get_vector_offset.
1717 (rs6000_split_vec_extract_var): Do not do the initial AND of
1718 element here, move the code to get_vector_offset.
1720 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1722 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1725 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1727 * config/rs6000/constraints.md: Improve documentation.
1729 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1732 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1733 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1735 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1737 * config.gcc: Remove "carrizo" support.
1738 * config/gcn/gcn-opts.h (processor_type): Likewise.
1739 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1740 * config/gcn/gcn.opt (gpu_type): Likewise.
1741 * config/gcn/t-omp-device: Likewise.
1743 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1746 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1747 * config/arm/arm.c (arm_gen_far_branch): New function
1749 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1751 2020-02-03 Julian Brown <julian@codesourcery.com>
1752 Tobias Burnus <tobias@codesourcery.com>
1754 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1756 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1759 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1760 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1763 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1765 PR rtl-optimization/91333
1766 * ira-color.c (struct allocno_color_data): Add member
1768 (init_allocno_threads): Set the member up.
1769 (bucket_allocno_compare_func): Add compare hard reg
1772 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1774 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1776 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1777 * config.in: Regenerated.
1778 * configure: Regenerated.
1779 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1780 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1781 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1783 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1785 * configure: Regenerate.
1787 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1789 PR rtl-optimization/91333
1790 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1791 reg preferences comparison up.
1793 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1795 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1796 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1797 aarch64-sve-builtins-base.h.
1798 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1799 aarch64-sve-builtins-base.cc.
1800 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1801 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1803 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1804 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1805 (svcvtnt): New functions.
1806 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1807 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1808 (svcvtnt): New functions.
1809 (svcvt): Add a form that converts f32 to bf16.
1810 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1811 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1813 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1814 Treat B as bfloat16_t.
1815 (ternary_bfloat_lane_base): New class.
1816 (ternary_bfloat_def): Likewise.
1817 (ternary_bfloat): New shape.
1818 (ternary_bfloat_lane_def): New class.
1819 (ternary_bfloat_lane): New shape.
1820 (ternary_bfloat_lanex2_def): New class.
1821 (ternary_bfloat_lanex2): New shape.
1822 (ternary_bfloat_opt_n_def): New class.
1823 (ternary_bfloat_opt_n): New shape.
1824 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1825 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1826 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1827 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1828 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1829 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1830 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1831 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1832 the pattern off the narrow mode instead of the wider one.
1833 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1834 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1835 (sve_fp_op): Handle them.
1836 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1837 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1839 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1841 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1842 * config/aarch64/aarch64-modes.def (BF): Move definition before
1843 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1844 (SVE_MODES): Handle BF modes.
1845 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1847 (aarch64_full_sve_mode): Likewise.
1848 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1850 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1851 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1852 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1853 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1855 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1857 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1858 (TYPES_all_data): Add bf16.
1859 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1860 (register_tuple_type): Increase buffer size.
1861 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1862 (bf16): New type suffix.
1863 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1864 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1865 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1866 Change type from all_data to all_arith.
1867 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1870 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1871 Matthew Malcomson <matthew.malcomson@arm.com>
1872 Richard Sandiford <richard.sandiford@arm.com>
1874 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1875 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1876 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1877 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1878 __ARM_FEATURE_MATMUL_FP64.
1879 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1880 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1881 be disabled at the same time.
1882 (f32mm): New extension.
1883 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1884 (AARCH64_FL_F64MM): Bump to the next bit up.
1885 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1886 (TARGET_SVE_F64MM): New macros.
1887 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1888 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1889 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1890 (UNSPEC_ZIP2Q): New unspeccs.
1891 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1892 (optab, sur, perm_insn): Handle the new unspecs.
1893 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1894 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1895 TARGET_SVE_F64MM instead of separate tests.
1896 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1897 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1898 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1899 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1900 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1901 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1902 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1903 (TYPES_s_signed): New macro.
1904 (TYPES_s_integer): Use it.
1905 (TYPES_d_float): New macro.
1906 (TYPES_d_data): Use it.
1907 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1908 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1909 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1910 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1911 (svmmla): New shape.
1912 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1913 template parameters.
1914 (ternary_resize2_lane_base): Likewise.
1915 (ternary_resize2_base): New class.
1916 (ternary_qq_lane_base): Likewise.
1917 (ternary_intq_uintq_lane_def): Likewise.
1918 (ternary_intq_uintq_lane): New shape.
1919 (ternary_intq_uintq_opt_n_def): New class
1920 (ternary_intq_uintq_opt_n): New shape.
1921 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1922 (ternary_uintq_intq_def): New class.
1923 (ternary_uintq_intq): New shape.
1924 (ternary_uintq_intq_lane_def): New class.
1925 (ternary_uintq_intq_lane): New shape.
1926 (ternary_uintq_intq_opt_n_def): New class.
1927 (ternary_uintq_intq_opt_n): New shape.
1928 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1929 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1930 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1931 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1933 (svdotprod_lane_impl): ...this new class.
1934 (svmmla_impl, svusdot_impl): New classes.
1935 (svdot_lane): Update to use svdotprod_lane_impl.
1936 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1937 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1939 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1940 function, with no types defined.
1941 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1942 AARCH64_FL_I8MM functions.
1943 (svmmla): New AARCH64_FL_F32MM function.
1944 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1945 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1946 AARCH64_FL_F64MM function.
1947 (REQUIRED_EXTENSIONS):
1949 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1951 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1954 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1956 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1957 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1958 (*movti_internal): Do not check for
1959 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1960 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1961 just after check for TARGET_AVX.
1962 (*movdf_internal): Ditto.
1963 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1964 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1965 * config/i386/sse.md (mov<mode>_internal): Only check
1966 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1967 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1968 (<sse>_andnot<mode>3<mask_name>): Move check for
1969 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1970 (<code><mode>3<mask_name>): Ditto.
1971 (*andnot<mode>3): Ditto.
1972 (*andnottf3): Ditto.
1973 (*<code><mode>3): Ditto.
1974 (*<code>tf3): Ditto.
1975 (*andnot<VI:mode>3): Remove
1976 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1977 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1978 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1979 (sse4_1_blendv<ssemodesuffix>): Ditto.
1980 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1981 Explain that tune applies to 128bit instructions only.
1983 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1985 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1986 to definition of hsa_kernel_description. Parse assembly to find SGPR
1987 and VGPR count of kernel and store in hsa_kernel_description.
1989 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1991 PR rtl-optimization/91838
1992 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1993 to truncate if allowed or reject combination.
1995 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1997 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1998 (find_inv_vars_cb): Likewise.
2000 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2002 * calls.c (special_function_p): Split out the check for DECL_NAME
2003 being non-NULL and fndecl being extern at file scope into a
2004 new maybe_special_function_p and call it. Drop check for fndecl
2005 being non-NULL that was after a usage of DECL_NAME (fndecl).
2006 * tree.h (maybe_special_function_p): New inline function.
2008 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2010 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2011 (mask_gather_load<mode>): ... here, and zero-initialize the
2013 (maskload<mode>di): Zero-initialize the destination.
2016 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2019 * doc/analyzer.texi (Limitations): Note that constraints on
2020 floating-point values are currently ignored.
2022 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2025 * symtab.c (symtab_node::noninterposable_alias): If localalias
2026 already exists, but is not usable, append numbers after it until
2027 a unique name is found. Formatting fix.
2030 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2033 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2035 * config/gcn/gcn.c (print_operand): Handle LTGT.
2036 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2038 2020-01-30 Richard Biener <rguenther@suse.de>
2040 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2041 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2043 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2045 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2046 without a DECL in .data.rel.ro.local.
2048 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2051 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2055 * config/i386/sse.md
2056 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2057 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2058 any_extend code iterator instead of always zero_extend.
2059 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2060 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2061 Use any_extend code iterator instead of always zero_extend.
2062 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2063 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2064 Use any_extend code iterator instead of always zero_extend.
2065 (*sse2_pmovmskb_ext): New define_insn.
2066 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2069 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2070 (*popcountsi2_zext_falsedep): New define_insn.
2072 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2074 * config.in: Regenerated.
2075 * configure: Regenerated.
2077 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2080 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2081 LLVM's assembler changed the default in version 9.
2083 2020-01-24 Jeff Law <law@redhat.com>
2085 PR tree-optimization/89689
2086 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2088 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2092 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2094 PR rtl-optimization/87763
2095 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2096 simplification to handle subregs as well as bare regs.
2097 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2099 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2102 * ira.c (ira): Revert use of simplified LRA algorithm.
2104 2020-01-29 Martin Jambor <mjambor@suse.cz>
2106 PR tree-optimization/92706
2107 * tree-sra.c (struct access): Fields first_link, last_link,
2108 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2109 next_rhs_queued and grp_rhs_queued respectively, new fields
2110 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2111 (struct assign_link): Field next renamed to next_rhs, new field
2112 next_lhs. Updated comment.
2113 (work_queue_head): Renamed to rhs_work_queue_head.
2114 (lhs_work_queue_head): New variable.
2115 (add_link_to_lhs): New function.
2116 (relink_to_new_repr): Also relink LHS lists.
2117 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2118 (add_access_to_lhs_work_queue): New function.
2119 (pop_access_from_work_queue): Renamed to
2120 pop_access_from_rhs_work_queue.
2121 (pop_access_from_lhs_work_queue): New function.
2122 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2124 (child_would_conflict_in_lacc): Renamed to
2125 child_would_conflict_in_acc. Adjusted parameter names.
2126 (create_artificial_child_access): New parameter set_grp_read, use it.
2127 (subtree_mark_written_and_enqueue): Renamed to
2128 subtree_mark_written_and_rhs_enqueue.
2129 (propagate_subaccesses_across_link): Renamed to
2130 propagate_subaccesses_from_rhs.
2131 (propagate_subaccesses_from_lhs): New function.
2132 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2135 2020-01-29 Martin Jambor <mjambor@suse.cz>
2137 PR tree-optimization/92706
2138 * tree-sra.c (struct access): Adjust comment of
2139 grp_total_scalarization.
2140 (find_access_in_subtree): Look for single children spanning an entire
2142 (scalarizable_type_p): Allow register accesses, adjust callers.
2143 (completely_scalarize): Remove function.
2144 (scalarize_elem): Likewise.
2145 (create_total_scalarization_access): Likewise.
2146 (sort_and_splice_var_accesses): Do not track total scalarization
2148 (analyze_access_subtree): New parameter totally, adjust to new meaning
2149 of grp_total_scalarization.
2150 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2151 (can_totally_scalarize_forest_p): New function.
2152 (create_total_scalarization_access): Likewise.
2153 (create_total_access_and_reshape): Likewise.
2154 (total_should_skip_creating_access): Likewise.
2155 (totally_scalarize_subtree): Likewise.
2156 (analyze_all_variable_accesses): Perform total scalarization after
2157 subaccess propagation using the new functions above.
2158 (initialize_constant_pool_replacements): Output initializers by
2159 traversing the access tree.
2161 2020-01-29 Martin Jambor <mjambor@suse.cz>
2163 * tree-sra.c (verify_sra_access_forest): New function.
2164 (verify_all_sra_access_forests): Likewise.
2165 (create_artificial_child_access): Set parent.
2166 (analyze_all_variable_accesses): Call the verifier.
2168 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2170 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2171 if called on indirect edge.
2172 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2173 speculative call if needed.
2175 2020-01-29 Richard Biener <rguenther@suse.de>
2177 PR tree-optimization/93428
2178 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2179 permutation when the load node is created.
2180 (vect_analyze_slp_instance): Re-use it here.
2182 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2184 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2186 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2188 PR rtl-optimization/93272
2189 * ira-lives.c (process_out_of_region_eh_regs): New function.
2190 (process_bb_node_lives): Call it.
2192 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2194 * coverage.c (read_counts_file): Make error message lowercase.
2196 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2198 * profile-count.c (profile_quality_display_names): Fix ordering.
2200 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2203 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2204 hash only when edge is first within the sequence.
2205 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2206 (symbol_table::create_edge): Do not set target_prob.
2207 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2209 (cgraph_edge::make_speculative): Drop target_prob parameter.
2210 (cgraph_edge::speculative_call_info): Remove.
2211 (cgraph_edge::first_speculative_call_target): New member function.
2212 (update_call_stmt_hash_for_removing_direct_edge): New function.
2213 (cgraph_edge::resolve_speculation): Rewrite to new API.
2214 (cgraph_edge::speculative_call_for_target): New member function.
2215 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2216 multiple speculation targets.
2217 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2219 (verify_speculative_call): Verify that targets form an interval.
2220 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2221 (cgraph_edge::first_speculative_call_target): New member function.
2222 (cgraph_edge::next_speculative_call_target): New member function.
2223 (cgraph_edge::speculative_call_target_ref): New member function.
2224 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2225 (cgraph_edge): Remove target_prob.
2226 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2227 Fix handling of speculative calls.
2228 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2229 * ipa-fnsummary.c (analyze_function_body): Likewise.
2230 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2231 * ipa-profile.c (dump_histogram): Fix formating.
2232 (ipa_profile_generate_summary): Watch for overflows.
2233 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2234 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2235 (update_indirect_edges_after_inlining): Update to new API.
2236 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2238 * profile-count.h: (profile_probability::adjusted): New.
2239 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2240 updating of profile.
2241 * value-prof.c (gimple_ic_transform): Rename to ...
2242 (dump_ic_profile): ... this one; update dumping.
2243 (stream_in_histogram_value): Fix formating.
2244 (gimple_value_profile_transformations): Update.
2246 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2249 * config/i386/i386.md (*movoi_internal_avx): Remove
2250 TARGET_SSE_TYPELESS_STORES check.
2251 (*movti_internal): Prefer TARGET_AVX over
2252 TARGET_SSE_TYPELESS_STORES.
2253 (*movtf_internal): Likewise.
2254 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2255 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2256 from TARGET_SSE_TYPELESS_STORES.
2258 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2260 * diagnostic-core.h (warning_at): Rename overload to...
2261 (warning_meta): ...this.
2262 (emit_diagnostic_valist): Delete decl of overload taking
2263 diagnostic_metadata.
2264 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2265 (warning_at): Rename overload taking diagnostic_metadata to...
2266 (warning_meta): ...this.
2268 2020-01-28 Richard Biener <rguenther@suse.de>
2270 PR tree-optimization/93439
2271 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2272 * tree-cfg.c (move_sese_region_to_fn): ... here.
2273 (verify_types_in_gimple_reference): Verify used cliques are
2276 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2279 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2280 argument of a pointer to struct gcc_options and pass it to
2281 parse_mtune_ctrl_str.
2282 (ix86_function_specific_restore): Pass opts to
2283 set_ix86_tune_features.
2284 (ix86_option_override_internal): Likewise.
2285 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2286 gcc_options and use it for x_ix86_tune_ctrl_string.
2288 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2290 PR rtl-optimization/87763
2291 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2292 simplification to handle subregs as well as bare regs.
2293 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2295 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2297 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2298 for reduction chains that (now) include a call.
2300 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2302 PR tree-optimization/92822
2303 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2304 out the don't-care elements of a vector whose significant elements
2305 are duplicates, make the don't-care elements duplicates too.
2307 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2309 PR tree-optimization/93434
2310 * tree-predcom.c (split_data_refs_to_components): Record which
2311 components have had aliasing loads removed. Prevent store-store
2312 commoning for all such components.
2314 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2317 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2318 -1 or is_vshift is true, use new_vector with number of elts npatterns
2319 rather than new_unary_operation.
2321 PR tree-optimization/93454
2322 * gimple-fold.c (fold_array_ctor_reference): Perform
2323 elt_size.to_uhwi () just once, instead of calling it in every
2324 iteration. Punt if that value is above size of the temporary
2325 buffer. Decrease third native_encode_expr argument when
2326 bufoff + elt_sz is above size of buf.
2328 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2330 * config/mips/mips.c (mips_declare_object_name)
2331 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2333 2020-01-27 Martin Liska <mliska@suse.cz>
2335 PR gcov-profile/93403
2336 * tree-profile.c (gimple_init_gcov_profiler): Generate
2337 both __gcov_indirect_call_profiler_v4 and
2338 __gcov_indirect_call_profiler_v4_atomic.
2340 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2343 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2345 (@aarch64_split_simd_mov<mode>): Use it.
2346 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2347 Leave the vec_extract patterns to handle 2-element vectors.
2348 (aarch64_simd_mov_from_<mode>high): Likewise.
2349 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2350 (vec_extractv2dfv1df): Likewise.
2352 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2354 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2355 jump conditions for *compare_condjump<GPI:mode>.
2357 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2360 * digraph.cc (test_edge::test_edge): Specify template for base
2363 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2365 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2367 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2369 * config/arc/arc-protos.h (gen_mlo): Remove.
2370 (gen_mhi): Likewise.
2371 * config/arc/arc.c (AUX_MULHI): Define.
2372 (arc_must_save_reister): Special handling for r58/59.
2373 (arc_compute_frame_size): Consider mlo/mhi registers.
2374 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2376 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2377 mlo/mhi name selection.
2378 (arc_restore_callee_saves): Don't early restore blink when ISR.
2379 (arc_expand_prologue): Add mlo/mhi saving.
2380 (arc_expand_epilogue): Add mlo/mhi restoring.
2383 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2384 numbering when MUL64 option is used.
2385 (DWARF2_FRAME_REG_OUT): Define.
2386 * config/arc/arc.md (arc600_stall): New pattern.
2387 (VUNSPEC_ARC_ARC600_STALL): Define.
2388 (mulsi64): Use correct mlo/mhi registers.
2389 (mulsi_600): Clean it up.
2390 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2392 (mhi_operand): Likewise.
2394 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2395 Petro Karashchenko <petro.karashchenko@ring.com>
2397 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2398 attributes if needed.
2399 (prepare_move_operands): Generate special unspec instruction for
2401 (arc_isuncached_mem_p): Propagate uncached attribute to each
2403 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2404 (VUNSPEC_ARC_STDI): Likewise.
2405 (ALLI): New mode iterator.
2406 (mALLI): New mode attribute.
2407 (lddi): New instruction pattern.
2409 (stdidi_split): Split instruction for architectures which are not
2410 supporting ll64 option.
2411 (lddidi_split): Likewise.
2413 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2415 PR rtl-optimization/92989
2416 * lra-lives.c (process_bb_lives): Update the live-in set before
2417 processing additional clobbers.
2419 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2421 PR rtl-optimization/93170
2422 * cselib.c (cselib_invalidate_regno_val): New function, split out
2424 (cselib_invalidate_regno): ...here.
2425 (cselib_invalidated_by_call_p): New function.
2426 (cselib_process_insn): Iterate over all the hard-register entries in
2427 REG_VALUES and invalidate any that cross call-clobbered registers.
2429 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2431 * dojump.c (split_comparison): Use HONOR_NANS rather than
2432 HONOR_SNANS when splitting LTGT.
2434 2020-01-27 Martin Liska <mliska@suse.cz>
2437 * opts.c (print_filtered_help): Exclude language-specific
2438 options from --help=common unless enabled in all FEs.
2440 2020-01-27 Martin Liska <mliska@suse.cz>
2442 * opts.c (print_help): Exclude params from
2443 all except --help=param.
2445 2020-01-27 Martin Liska <mliska@suse.cz>
2448 * config/i386/i386-features.c (make_resolver_func):
2449 Align the code with ppc64 target implementation.
2450 Do not generate a unique name for resolver function.
2452 2020-01-27 Richard Biener <rguenther@suse.de>
2454 PR tree-optimization/93397
2455 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2456 converted reduction chain SLP graph adjustment.
2458 2020-01-26 Marek Polacek <polacek@redhat.com>
2461 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2464 2020-01-26 Jason Merrill <jason@redhat.com>
2467 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2470 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2472 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2473 (rx_setmem): Likewise.
2475 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2478 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2479 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2480 drop <di> from constraint of last operand.
2483 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2484 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2485 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2487 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2490 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2493 2020-01-24 Jeff Law <law@redhat.com>
2495 PR tree-optimization/92788
2496 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2499 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2502 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2503 *avx_vperm_broadcast_<mode>,
2504 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2505 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2506 Move before avx2_perm<mode>/avx512f_perm<mode>.
2509 * simplify-rtx.c (simplify_const_unary_operation,
2510 simplify_const_binary_operation): Punt for mode precision above
2511 MAX_BITSIZE_MODE_ANY_INT.
2513 2020-01-24 Andrew Pinski <apinski@marvell.com>
2515 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2518 2020-01-24 Jeff Law <law@redhat.com>
2521 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2522 for REGs. Call output_operand_lossage to get more reasonable
2525 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2527 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2528 gcn_fp_compare_operator.
2529 (vec_cmpu<mode>di): Use gcn_compare_operator.
2530 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2531 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2532 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2533 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2534 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2535 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2536 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2537 gcn_fp_compare_operator.
2538 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2539 gcn_fp_compare_operator.
2540 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2541 gcn_fp_compare_operator.
2542 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2543 gcn_fp_compare_operator.
2545 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2547 * doc/install.texi (Cross-Compiler-Specific Options): Document
2548 `--with-toolexeclibdir' option.
2550 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2552 * target.def (flags_regnum): Also mention effect on delay slot filling.
2553 * doc/tm.texi: Regenerate.
2555 2020-01-23 Jeff Law <law@redhat.com>
2557 PR translation/90162
2558 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2560 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2563 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2566 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2568 PR rtl-optimization/93402
2569 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2572 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2574 * config.in: Regenerated.
2575 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2576 for TARGET_LIBC_GNUSTACK.
2577 * configure: Regenerated.
2578 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2579 found to be 2.31 or greater.
2581 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2583 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2585 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2586 (mips_asm_file_end): New function. Delegate to
2587 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2588 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2590 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2593 * config/i386/i386-modes.def (POImode): New mode.
2594 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2595 * config/i386/i386.md (DPWI): New mode attribute.
2596 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2598 (QPWI): ... this. Use POI instead of OI for TImode.
2599 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2600 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2603 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2606 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2608 (speculation_tracker_rev): New pattern.
2609 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2610 Use speculation_tracker_rev to track the inverse condition.
2612 2020-01-23 Richard Biener <rguenther@suse.de>
2614 PR tree-optimization/93381
2615 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2616 alias-set of the def as argument and record the first one.
2617 (vn_walk_cb_data::first_set): New member.
2618 (vn_reference_lookup_3): Pass the alias-set of the current def
2619 to push_partial_def. Fix alias-set used in the aggregate copy
2621 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2622 * real.c (clear_significand_below): Fix out-of-bound access.
2624 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2627 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2628 New define_insn patterns.
2630 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2632 * doc/sourcebuild.texi (check-function-bodies): Add an
2633 optional target/xfail selector.
2635 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2637 PR rtl-optimization/93124
2638 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2639 bare USE and CLOBBER insns.
2641 2020-01-22 Andrew Pinski <apinski@marvell.com>
2643 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2645 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2648 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2649 diagnostic_manager into "ana" namespace.
2650 * selftest-run-tests.c (selftest::run_tests): Update for move of
2651 selftest::run_analyzer_selftests to
2652 ana::selftest::run_analyzer_selftests.
2654 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2656 * cfgexpand.c (union_stack_vars): Update the size.
2658 2020-01-22 Richard Biener <rguenther@suse.de>
2660 PR tree-optimization/93381
2661 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2662 throughout, handle all conversions the same.
2664 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2667 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2668 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2669 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2670 Call force_reg on high_in2 unconditionally.
2672 2020-01-22 Martin Liska <mliska@suse.cz>
2674 PR tree-optimization/92924
2675 * profile.c (compute_value_histograms): Divide
2678 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2681 * output.h (assemble_name_resolve): Declare.
2682 * varasm.c (assemble_name_resolve): New function.
2683 (assemble_name): Use it.
2684 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2686 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2688 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2689 update_web_docs_git instead of update_web_docs_svn.
2691 2020-01-21 Andrew Pinski <apinski@marvell.com>
2694 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2695 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2696 (*tlsgd_small_<mode>): Likewise.
2697 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2698 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2699 register. Convert that register back to dest using convert_mode.
2701 2020-01-21 Jim Wilson <jimw@sifive.com>
2703 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2706 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2707 Uros Bizjak <ubizjak@gmail.com>
2710 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2712 (legitimize_tls_address): Do GNU2 TLS address computation in
2713 ptr_mode and zero-extend result to Pmode.
2714 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2715 :P with :PTR and Pmode with ptr_mode.
2716 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2717 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2718 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2720 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2723 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2724 the last two operands are CONST_INT_P before using them as such.
2726 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2728 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2729 to get the integer element types.
2731 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2733 * config/aarch64/aarch64-sve-builtins.h
2734 (function_expander::convert_to_pmode): Declare.
2735 * config/aarch64/aarch64-sve-builtins.cc
2736 (function_expander::convert_to_pmode): New function.
2737 (function_expander::get_contiguous_base): Use it.
2738 (function_expander::prepare_gather_address_operands): Likewise.
2739 * config/aarch64/aarch64-sve-builtins-sve2.cc
2740 (svwhilerw_svwhilewr_impl::expand): Likewise.
2742 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2745 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2746 cfun->machine->label_is_assembled.
2747 (aarch64_print_patchable_function_entry): New.
2748 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2749 * config/aarch64/aarch64.h (struct machine_function): New field,
2752 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2755 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2758 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2761 * cgraph.c (cgraph_edge::resolve_speculation,
2762 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2763 call_stmt_site_hash.
2765 2020-01-21 Martin Liska <mliska@suse.cz>
2767 * config/rs6000/rs6000.c (common_mode_defined): Remove
2770 2020-01-21 Richard Biener <rguenther@suse.de>
2772 PR tree-optimization/92328
2773 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2774 type when value-numbering same-sized store by inserting a
2776 (eliminate_dom_walker::eliminate_stmt): When eliminating
2777 a redundant store handle bit-reinterpretation of the same value.
2779 2020-01-21 Andrew Pinski <apinski@marvel.com>
2782 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2784 (prepare_block_for_update): This. Use a worklist instead of
2787 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2789 * gcc/config/arm/arm.c (clear_operation_p):
2790 Initialise last_regno, skip first iteration
2791 based on the first_set value and use ints instead
2792 of the unnecessary HOST_WIDE_INTs.
2794 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2797 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2798 compare_mode other than SFmode or DFmode.
2800 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2803 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2804 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2805 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2807 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2809 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2811 2020-01-20 Andrew Pinski <apinski@marvell.com>
2814 * targhooks.c (default_print_patchable_function_entry): Use
2815 output_asm_insn to emit the nop instruction.
2817 2020-01-20 Fangrui Song <maskray@google.com>
2820 * targhooks.c (default_print_patchable_function_entry): Align to
2823 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2826 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2827 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2828 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2829 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2830 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2831 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2832 Remove the {q} suffix from lea.
2833 (*tls_dynamic_gnu2_call_64): Renamed to ...
2834 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2835 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2836 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2837 Pass Pmode to gen_tls_dynamic_gnu2_64.
2839 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2841 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2843 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2845 * config/aarch64/aarch64-sve-builtins-base.cc
2846 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2848 2020-01-20 Richard Biener <rguenther@suse.de>
2851 * dwarf2out.c (prune_unused_types): Unconditionally mark
2852 called function DIEs.
2854 2020-01-20 Martin Liska <mliska@suse.cz>
2856 PR tree-optimization/93199
2857 * tree-eh.c (struct leh_state): Add
2858 new field outer_non_cleanup.
2859 (cleanup_is_dead_in): Pass leh_state instead
2860 of eh_region. Add a checking that state->outer_non_cleanup
2861 points to outer non-clean up region.
2862 (lower_try_finally): Record outer_non_cleanup
2864 (lower_catch): Likewise.
2865 (lower_eh_filter): Likewise.
2866 (lower_eh_must_not_throw): Likewise.
2867 (lower_cleanup): Likewise.
2869 2020-01-20 Richard Biener <rguenther@suse.de>
2871 PR tree-optimization/93094
2872 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2873 (vect_transform_loop): Likewise.
2874 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2875 loop_vectorized_call to vect_transform_loop.
2876 * tree-vect-loop.c (vect_transform_loop): Pass down
2877 loop_vectorized_call to vect_loop_versioning.
2878 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2879 the earlier discovered loop_vectorized_call.
2881 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2883 * doc/contribute.texi: Update for SVN -> Git transition.
2884 * doc/install.texi: Likewise.
2886 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2889 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2890 speculative targets.
2891 (verify_speculative_call): New function
2892 (cgraph_node::verify_node): Use it.
2893 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2896 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2899 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2900 (cgraph_edge::make_direct): Remove all indirect targets.
2901 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2902 (cgraph_node::verify_node): Verify that only one call_stmt or
2903 lto_stmt_uid is set.
2904 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2906 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2907 (lto_output_ref): Simplify streaming of stmt.
2908 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2910 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2912 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2913 Mark parameter unused.
2915 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2917 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2919 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2921 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2923 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2925 * Makefile.in: Add coroutine-passes.o.
2926 * builtin-types.def (BT_CONST_SIZE): New.
2927 (BT_FN_BOOL_PTR): New.
2928 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2929 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2930 * coroutine-builtins.def: New file.
2931 * coroutine-passes.cc: New file.
2932 * function.h (struct GTY function): Add a bit to indicate that the
2933 function is a coroutine component.
2934 * internal-fn.c (expand_CO_FRAME): New.
2935 (expand_CO_YIELD): New.
2936 (expand_CO_SUSPN): New.
2937 (expand_CO_ACTOR): New.
2938 * internal-fn.def (CO_ACTOR): New.
2942 * passes.def: Add pass_coroutine_lower_builtins,
2943 pass_coroutine_early_expand_ifns.
2944 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2945 (make_pass_coroutine_early_expand_ifns): New.
2946 * doc/invoke.texi: Document the fcoroutines command line
2949 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2951 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2954 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2955 after checking the argument is a REG. Don't use REGNO (reg)
2956 again to set last_regno, reuse regno variable instead.
2958 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2960 * doc/analyzer.texi (Limitations): Add note about NaN.
2962 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2963 Sudakshina Das <sudi.das@arm.com>
2965 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2966 and valid immediate.
2967 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2968 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2969 * config/arm/constraints.md (Pg): New.
2970 * config/arm/predicates.md (long_shift_imm): New.
2971 (arm_reg_or_long_shift_imm): Likewise.
2972 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2973 (thumb2_lsll): Likewise.
2976 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2977 Sudakshina Das <sudi.das@arm.com>
2979 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2980 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2981 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2982 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2983 * config/arm/thumb2.md (thumb2_asrl): New.
2984 (thumb2_lsll): Likewise.
2986 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2988 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2991 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2993 * gdbinit.in (help-gcc-hooks): New command.
2994 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2995 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2998 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3000 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3001 correct target macro.
3003 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3005 * config/aarch64/aarch64-protos.h
3006 (aarch64_sve_ld1ro_operand_p): New.
3007 * config/aarch64/aarch64-sve-builtins-base.cc
3008 (class load_replicate): New.
3009 (class svld1ro_impl): New.
3010 (class svld1rq_impl): Change to inherit from load_replicate.
3011 (svld1ro): New sve intrinsic function base.
3012 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3013 New DEF_SVE_FUNCTION.
3014 * config/aarch64/aarch64-sve-builtins-base.h
3015 (svld1ro): New decl.
3016 * config/aarch64/aarch64-sve-builtins.cc
3017 (function_expander::add_mem_operand): Modify assert to allow
3019 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3021 * config/aarch64/aarch64.c
3022 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3023 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3024 (aarch64_sve_ld1ro_operand_p): New.
3025 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3026 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3027 * config/aarch64/predicates.md
3028 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3030 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3032 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3033 Introduce this ACLE specified predefined macro.
3034 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3035 (fp): Disabling this disables f64mm.
3036 (simd): Disabling this disables f64mm.
3037 (fp16): Disabling this disables f64mm.
3038 (sve): Disabling this disables f64mm.
3039 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3040 (AARCH64_ISA_F64MM): New.
3041 (TARGET_F64MM): New.
3042 * doc/invoke.texi (f64mm): Document new option.
3044 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3046 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3047 (neoversen1_tunings): Likewise.
3049 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3052 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3053 Add assert to ensure prolog has been emitted.
3054 (aarch64_split_atomic_op): Likewise.
3055 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3056 Use epilogue_completed rather than reload_completed.
3057 (aarch64_atomic_exchange<mode>): Likewise.
3058 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3059 (atomic_nand<mode>): Likewise.
3060 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3061 (atomic_fetch_nand<mode>): Likewise.
3062 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3063 (atomic_nand_fetch<mode>): Likewise.
3065 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3068 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3070 (REVERSE_CONDITION): Delete.
3071 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3072 (CCFP_CCFPE): Likewise.
3073 (e): New mode attribute.
3074 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3075 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3076 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3077 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3078 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3079 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3080 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3081 name of generator from gen_ccmpdi to gen_ccmpccdi.
3082 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3083 the previous comparison but aren't able to, use the new ccmp_rev
3086 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3088 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3089 than testing directly for INTEGER_CST.
3090 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3092 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3094 PR tree-optimization/93292
3095 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3096 get_vectype_for_scalar_type returns NULL.
3098 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3100 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3101 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3103 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3105 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3107 * params.opt: (max-predicted-iterations): Set bounds.
3108 * predict.c (real_almost_one, real_br_prob_base,
3109 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3110 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3111 probabilities; do not truncate to reg_br_prob_bases.
3112 (estimate_loops_at_level): Pass max_cyclic_prob.
3113 (estimate_loops): Compute max_cyclic_prob.
3114 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3116 * profile-count.c (profile_probability::to_sreal): New.
3117 * profile-count.h (class sreal): Move up in file.
3118 (profile_probability::to_sreal): Declare.
3120 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3123 (arm_invalid_conversion): New function for target hook.
3124 (arm_invalid_unary_op): New function for target hook.
3125 (arm_invalid_binary_op): New function for target hook.
3127 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3129 * config.gcc: Add arm_bf16.h.
3130 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3131 (arm_simd_builtin_std_type): Add BFmode.
3132 (arm_init_simd_builtin_types): Define element types for vector types.
3133 (arm_init_bf16_types): New function.
3134 (arm_init_builtins): Add arm_init_bf16_types function call.
3135 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3136 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3137 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3138 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3139 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3140 (arm_mangle_type): Add __bf16.
3141 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3142 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3143 arm_bf16_ptr_type_node.
3144 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3145 define_split between ARM registers.
3146 * config/arm/arm_bf16.h: New file.
3147 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3148 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3150 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3151 * config/arm/vfp.md: Add BFmode to movhf patterns.
3153 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3154 Andre Vieira <andre.simoesdiasvieira@arm.com>
3156 * config/arm/arm-cpus.in (mve, mve_float): New features.
3157 (dsp, mve, mve.fp): New options.
3158 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3159 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3160 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3162 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3163 Thomas Preud'homme <thomas.preudhomme@arm.com>
3165 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3167 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3168 error for using -mcmse when targeting Armv8.1-M Mainline.
3170 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3171 Thomas Preud'homme <thomas.preudhomme@arm.com>
3173 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3174 address in r4 when targeting Armv8.1-M Mainline.
3175 (nonsecure_call_value_internal): Likewise.
3176 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3177 a register match_operand again. Emit BLXNS when targeting
3179 (nonsecure_call_value_reg_thumb2): Likewise.
3181 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3182 Thomas Preud'homme <thomas.preudhomme@arm.com>
3184 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3185 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3186 variable as true when floating-point ABI is not hard. Replace
3187 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3188 Generate VLSTM and VLLDM instruction respectively before and
3189 after a function call to cmse_nonsecure_call function.
3190 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3191 (VUNSPEC_VLLDM): Likewise.
3192 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3193 (lazy_load_multiple_insn): Likewise.
3195 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3196 Thomas Preud'homme <thomas.preudhomme@arm.com>
3198 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3199 (arm_emit_vfp_multi_reg_pop): Likewise.
3200 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3201 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3202 restore callee-saved VFP registers.
3204 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3205 Thomas Preud'homme <thomas.preudhomme@arm.com>
3207 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3208 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3209 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3210 callee-saved GPRs as well as clear ip register before doing a nonsecure
3211 call then restore callee-saved GPRs after it when targeting
3213 (arm_reorg): Adapt to function rename.
3215 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3216 Thomas Preud'homme <thomas.preudhomme@arm.com>
3218 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3219 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3220 clear_vfp_multiple pattern based on a new vfp parameter.
3221 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3222 targeting Armv8.1-M Mainline.
3223 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3224 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3225 whether VFP registers are available before looking call_used_regs for a
3227 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3228 of prototype of clear_operation_p.
3229 (clear_vfp_multiple_operation): New predicate.
3230 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3231 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3233 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3234 Thomas Preud'homme <thomas.preudhomme@arm.com>
3236 * config/arm/arm-protos.h (clear_operation_p): Declare.
3237 * config/arm/arm.c (clear_operation_p): New function.
3238 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3239 targeting Armv8.1-M Mainline or successor.
3240 (output_return_instruction): Only output APSR register clearing if
3241 Armv8.1-M Mainline instructions not available.
3242 (thumb_exit): Likewise.
3243 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3244 * config/arm/thumb2.md (clear_apsr): New define_insn.
3245 (clear_multiple): Likewise.
3246 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3248 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3249 Thomas Preud'homme <thomas.preudhomme@arm.com>
3251 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3252 (use_return_insn): Also return false for Armv8.1-M Mainline.
3253 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3254 Mainline instructions are available.
3255 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3256 when targeting Armv8.1-M Mainline Security Extensions.
3257 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3258 Mainline entry function.
3259 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3260 targeting Armv8.1-M Mainline or successor.
3261 (arm_expand_epilogue): Fix indentation of caller-saved register
3262 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3264 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3265 (FP_SYSREGS): Likewise.
3266 (enum vfp_sysregs_encoding): Define enum.
3267 (fp_sysreg_names): Declare.
3268 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3269 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3270 (pop_fpsysreg_insn): Likewise.
3272 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3273 Thomas Preud'homme <thomas.preudhomme@arm.com>
3275 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3276 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3277 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3278 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3279 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3280 (ARMv8_1m_main): New feature group.
3281 (armv8.1-m.main): New architecture.
3282 * config/arm/arm-tables.opt: Regenerate.
3283 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3284 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3285 (arm_options_perform_arch_sanity_checks): Error out when targeting
3286 Armv8.1-M Mainline Security Extensions.
3287 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3289 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3291 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3292 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3293 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3294 aarch64_bfdot_laneq): New.
3295 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3296 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3297 vbfdotq_laneq_f32): New.
3298 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3299 VBFMLA_W, VBF): New.
3300 (isquadop): Add V4BF, V8BF.
3302 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3304 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3305 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3306 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3307 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3308 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3309 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3310 usdot_laneq, sudot_lane,sudot_laneq): New.
3311 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3312 (aarch64_<sur>dot_lane): New.
3313 * config/aarch64/arm_neon.h (vusdot_s32): New.
3315 (vusdot_lane_s32): New.
3316 (vsudot_lane_s32): New.
3317 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3318 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3320 2020-01-16 Martin Liska <mliska@suse.cz>
3322 * value-prof.c (dump_histogram_value): Fix
3323 obvious spacing issue.
3325 2020-01-16 Andrew Pinski <apinski@marvell.com>
3327 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3328 !storage_order_barrier_p.
3330 2020-01-16 Andrew Pinski <apinski@marvell.com>
3332 * sched-int.h (_dep): Add unused bit-field field for the padding.
3333 * sched-deps.c (init_dep_1): Init unused field.
3335 2020-01-16 Andrew Pinski <apinski@marvell.com>
3337 * optabs.h (create_expand_operand): Initialize target field also.
3339 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3341 PR tree-optimization/92429
3342 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3343 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3345 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3348 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3350 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3351 aarch64_sve_int_mode to each mode.
3353 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3355 * doc/analyzer.texi (Overview): Add note about
3356 -fdump-ipa-analyzer.
3358 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3360 PR tree-optimization/93231
3361 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3362 input_type is unsigned. Use tree_to_shwi for shift constant.
3363 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3364 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3365 inputs more efficiently.
3367 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3369 * config/i386/i386.md (*movsf_internal): Do not require
3370 SSE2 ISA for alternatives 14 and 15.
3372 2020-01-15 Richard Biener <rguenther@suse.de>
3375 * tree-eh.c (sink_clobbers): If we already visited the destination
3376 block do not defer insertion.
3377 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3378 the purpose of defered insertion.
3380 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3382 * BASE-VER: Bump to 10.0.1.
3384 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3386 PR tree-optimization/93247
3387 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3388 type of the stmt that we're going to vectorize.
3390 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3392 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3393 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3396 2020-01-15 Martin Liska <mliska@suse.cz>
3398 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3399 2 calls of streamer_read_hwi in a function call.
3401 2020-01-15 Richard Biener <rguenther@suse.de>
3403 * alias.c (record_alias_subset): Avoid redundant work when
3404 subset is already recorded.
3406 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3408 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3409 the analyzer options provide CWE identifiers.
3411 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3413 * tree-diagnostic-path.cc (path_summary::event_range::print):
3414 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3415 using get_pure_location.
3417 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3419 PR tree-optimization/93262
3420 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3421 perform head trimming only if the last argument is constant,
3422 either all ones, or larger or equal to head trim, in the latter
3423 case decrease the last argument by head_trim.
3425 PR tree-optimization/93249
3426 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3427 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3428 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3429 perform head trim unless we can prove there are no '\0' chars
3430 from the source among the first head_trim chars.
3432 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3434 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3436 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3439 * config/i386/sse.md
3440 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3441 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3442 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3443 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3444 just a single alternative instead of two, make operands 1 and 2
3447 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3450 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3453 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3455 * Makefile.in (lang_opt_files): Add analyzer.opt.
3456 (ANALYZER_OBJS): New.
3457 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3458 tristate.o and ANALYZER_OBJS.
3459 (TEXI_GCCINT_FILES): Add analyzer.texi.
3460 * common.opt (-fanalyzer): New driver option.
3461 * config.in: Regenerate.
3462 * configure: Regenerate.
3463 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3464 (gccdepdir): Also create depdir for "analyzer" subdir.
3465 * digraph.cc: New file.
3466 * digraph.h: New file.
3467 * doc/analyzer.texi: New file.
3468 * doc/gccint.texi ("Static Analyzer") New menu item.
3469 (analyzer.texi): Include it.
3470 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3471 ("Warning Options"): Add static analysis warnings to the list.
3472 (-Wno-analyzer-double-fclose): New option.
3473 (-Wno-analyzer-double-free): New option.
3474 (-Wno-analyzer-exposure-through-output-file): New option.
3475 (-Wno-analyzer-file-leak): New option.
3476 (-Wno-analyzer-free-of-non-heap): New option.
3477 (-Wno-analyzer-malloc-leak): New option.
3478 (-Wno-analyzer-possible-null-argument): New option.
3479 (-Wno-analyzer-possible-null-dereference): New option.
3480 (-Wno-analyzer-null-argument): New option.
3481 (-Wno-analyzer-null-dereference): New option.
3482 (-Wno-analyzer-stale-setjmp-buffer): New option.
3483 (-Wno-analyzer-tainted-array-index): New option.
3484 (-Wno-analyzer-use-after-free): New option.
3485 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3486 (-Wno-analyzer-use-of-uninitialized-value): New option.
3487 (-Wanalyzer-too-complex): New option.
3488 (-fanalyzer-call-summaries): New warning.
3489 (-fanalyzer-checker=): New warning.
3490 (-fanalyzer-fine-grained): New warning.
3491 (-fno-analyzer-state-merge): New warning.
3492 (-fno-analyzer-state-purge): New warning.
3493 (-fanalyzer-transitivity): New warning.
3494 (-fanalyzer-verbose-edges): New warning.
3495 (-fanalyzer-verbose-state-changes): New warning.
3496 (-fanalyzer-verbosity=): New warning.
3497 (-fdump-analyzer): New warning.
3498 (-fdump-analyzer-callgraph): New warning.
3499 (-fdump-analyzer-exploded-graph): New warning.
3500 (-fdump-analyzer-exploded-nodes): New warning.
3501 (-fdump-analyzer-exploded-nodes-2): New warning.
3502 (-fdump-analyzer-exploded-nodes-3): New warning.
3503 (-fdump-analyzer-supergraph): New warning.
3504 * doc/sourcebuild.texi (dg-require-dot): New.
3505 (dg-check-dot): New.
3506 * gdbinit.in (break-on-saved-diagnostic): New command.
3507 * graphviz.cc: New file.
3508 * graphviz.h: New file.
3509 * ordered-hash-map-tests.cc: New file.
3510 * ordered-hash-map.h: New file.
3511 * passes.def (pass_analyzer): Add before
3512 pass_ipa_whole_program_visibility.
3513 * selftest-run-tests.c (selftest::run_tests): Call
3514 selftest::ordered_hash_map_tests_cc_tests.
3515 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3517 * shortest-paths.h: New file.
3518 * timevar.def (TV_ANALYZER): New timevar.
3519 (TV_ANALYZER_SUPERGRAPH): Likewise.
3520 (TV_ANALYZER_STATE_PURGE): Likewise.
3521 (TV_ANALYZER_PLAN): Likewise.
3522 (TV_ANALYZER_SCC): Likewise.
3523 (TV_ANALYZER_WORKLIST): Likewise.
3524 (TV_ANALYZER_DUMP): Likewise.
3525 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3526 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3527 * tree-pass.h (make_pass_analyzer): New decl.
3528 * tristate.cc: New file.
3529 * tristate.h: New file.
3531 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3534 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3535 alternatives 9 and 10.
3537 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3539 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3540 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3541 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3542 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3543 (selftest::hash_map_tests_c_tests): Call it.
3544 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3545 New static constant, using the value of = H::empty_zero_p.
3546 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3547 from default_hash_traits <Value>.
3548 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3550 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3551 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3552 calls to mark_empty with !Descriptor::empty_zero_p.
3553 (hash_table::empty_slow): Conditionalize the memset call with a
3554 check that Descriptor::empty_zero_p; otherwise, loop through the
3555 entries calling mark_empty on them.
3556 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3557 (pointer_hash::empty_zero_p): Likewise.
3558 (pair_hash::empty_zero_p): Likewise.
3559 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3561 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3562 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3563 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3564 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3565 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3566 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3567 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3568 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3570 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3573 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3575 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3576 fix typo on return value.
3578 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3581 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3583 (cgraph_edge::make_speculative): Add param for setting speculative_id
3585 (cgraph_edge::speculative_call_info): Update comments and find reference
3586 by speculative_id for multiple indirect targets.
3587 (cgraph_edge::resolve_speculation): Decrease the speculations
3588 for indirect edge, drop it's speculative if not direct target
3589 left. Update comments.
3590 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3591 (cgraph_node::dump): Print num_speculative_call_targets.
3592 (cgraph_node::verify_node): Don't report error if speculative
3593 edge not include statement.
3594 (cgraph_edge::num_speculative_call_targets_p): New function.
3595 * cgraph.h (int common_target_id): Remove.
3596 (int common_target_probability): Remove.
3597 (num_speculative_call_targets): New variable.
3598 (make_speculative): Add param for setting speculative_id.
3599 (cgraph_edge::num_speculative_call_targets_p): New declare.
3600 (target_prob): New variable.
3601 (speculative_id): New variable.
3602 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3603 call summaries for multiple speculative call targets.
3604 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3605 * ipa-profile.c (struct speculative_call_target): New struct.
3606 (class speculative_call_summary): New class.
3607 (class speculative_call_summaries): New class.
3608 (call_sums): New variable.
3609 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3610 (ipa_profile_write_edge_summary): New function.
3611 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3612 (ipa_profile_dump_all_summaries): New function.
3613 (ipa_profile_read_edge_summary): New function.
3614 (ipa_profile_read_summary_section): New function.
3615 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3616 (ipa_profile): Generate num_speculative_call_targets from
3618 * ipa-ref.h (speculative_id): New variable.
3619 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3620 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3621 common_target_probability. Stream out speculative_id and
3622 num_speculative_call_targets.
3623 (input_edge): Likewise.
3624 * predict.c (dump_prediction): Remove edges count assert to be
3626 * symtab.c (symtab_node::create_reference): Init speculative_id.
3627 (symtab_node::clone_references): Clone speculative_id.
3628 (symtab_node::clone_referring): Clone speculative_id.
3629 (symtab_node::clone_reference): Clone speculative_id.
3630 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3631 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3632 if indirect call contains multiple speculative targets.
3633 * value-prof.h (check_ic_target): Remove.
3634 * value-prof.c (gimple_value_profile_transformations):
3635 Use void function gimple_ic_transform.
3636 * value-prof.c (gimple_ic_transform): Handle topn case.
3637 Fix comment typos. Change it to a void function.
3639 2020-01-13 Andrew Pinski <apinski@marvell.com>
3641 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3642 (octeontx2t98): New define.
3643 (octeontx2t96): New define.
3644 (octeontx2t93): New define.
3645 (octeontx2f95): New define.
3646 (octeontx2f95n): New define.
3647 (octeontx2f95mm): New define.
3648 * config/aarch64/aarch64-tune.md: Regenerate.
3649 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3651 2020-01-13 Jason Merrill <jason@redhat.com>
3653 PR c++/33799 - destroy return value if local cleanup throws.
3654 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3656 2020-01-13 Martin Liska <mliska@suse.cz>
3658 * ipa-cp.c (get_max_overall_size): Use newly
3659 renamed param param_ipa_cp_unit_growth.
3660 * params.opt: Remove legacy param name.
3662 2020-01-13 Martin Sebor <msebor@redhat.com>
3664 PR tree-optimization/93213
3665 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3666 stores to be eliminated.
3668 2020-01-13 Martin Liska <mliska@suse.cz>
3670 * opts.c (print_help): Do not print CL_PARAM
3671 and CL_WARNING for CL_OPTIMIZATION.
3673 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3676 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3677 depending on optimization settings.
3679 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3681 PR tree-optimization/90838
3682 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3683 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3684 argument rather than to initialize temporary for targets that
3685 don't use the mode argument at all. Initialize ctzval to avoid
3688 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3690 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3691 * tree-core.h: Document it.
3692 * gimplify.c (gimplify_omp_workshare): Set it.
3693 * omp-low.c (lower_omp_target): Use it.
3694 * tree-pretty-print.c (dump_omp_clause): Print it.
3696 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3697 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3699 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3701 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3702 * common.opt (fdiagnostics-path-format=): New option.
3703 (diagnostic_path_format): New enum.
3704 (fdiagnostics-show-path-depths): New option.
3705 * coretypes.h (diagnostic_event_id_t): New forward decl.
3706 * diagnostic-color.c (color_dict): Add "path".
3707 * diagnostic-event-id.h: New file.
3708 * diagnostic-format-json.cc (json_from_expanded_location): Make
3710 (json_end_diagnostic): Call context->make_json_for_path if it
3711 exists and the diagnostic has a path.
3712 (diagnostic_output_format_init): Clear context->print_path.
3713 * diagnostic-path.h: New file.
3714 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3715 when printing a run of events in a diagnostic_path so that they
3716 all get the same color.
3717 (layout::m_diagnostic_path_p): New field.
3718 (layout::layout): Initialize it.
3719 (layout::print_any_labels): Don't colorize the label text for an
3720 event in a diagnostic_path.
3721 (gcc_rich_location::add_location_if_nearby): Add
3722 "restrict_to_current_line_spans" and "label" params. Pass the
3723 former to layout.maybe_add_location_range; pass the latter
3724 when calling add_range.
3725 * diagnostic.c: Include "diagnostic-path.h".
3726 (diagnostic_initialize): Initialize context->path_format and
3727 context->show_path_depths.
3728 (diagnostic_show_any_path): New function.
3729 (diagnostic_path::interprocedural_p): New function.
3730 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3731 (simple_diagnostic_path::num_events): New function.
3732 (simple_diagnostic_path::get_event): New function.
3733 (simple_diagnostic_path::add_event): New function.
3734 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3735 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3736 (debug): New overload taking a diagnostic_path *.
3737 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3738 * diagnostic.h (enum diagnostic_path_format): New enum.
3739 (json::value): New forward decl.
3740 (diagnostic_context::path_format): New field.
3741 (diagnostic_context::show_path_depths): New field.
3742 (diagnostic_context::print_path): New callback field.
3743 (diagnostic_context::make_json_for_path): New callback field.
3744 (diagnostic_show_any_path): New decl.
3745 (json_from_expanded_location): New decl.
3746 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3747 (-fdiagnostics-show-path-depths): New option.
3748 (-fdiagnostics-color): Add "path" to description of default
3749 GCC_COLORS; describe it.
3750 (-fdiagnostics-format=json): Document how diagnostic paths are
3751 represented in the JSON output format.
3752 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3753 Add optional params "restrict_to_current_line_spans" and "label".
3754 * opts.c (common_handle_option): Handle
3755 OPT_fdiagnostics_path_format_ and
3756 OPT_fdiagnostics_show_path_depths.
3757 * pretty-print.c: Include "diagnostic-event-id.h".
3758 (pp_format): Implement "%@" format code for printing
3759 diagnostic_event_id_t *.
3760 (selftest::test_pp_format): Add tests for "%@".
3761 * selftest-run-tests.c (selftest::run_tests): Call
3762 selftest::tree_diagnostic_path_cc_tests.
3763 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3764 * toplev.c (general_init): Initialize global_dc->path_format and
3765 global_dc->show_path_depths.
3766 * tree-diagnostic-path.cc: New file.
3767 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3768 non-static. Drop "diagnostic" param in favor of storing the
3769 original value of "where" and re-using it.
3770 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3771 maybe_unwind_expanded_macro_loc.
3772 (tree_diagnostics_defaults): Initialize context->print_path and
3773 context->make_json_for_path.
3774 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3776 (default_tree_make_json_for_path): New decl.
3777 (maybe_unwind_expanded_macro_loc): New decl.
3779 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3781 PR tree-optimization/93210
3782 * fold-const.h (native_encode_initializer,
3783 can_native_interpret_type_p): Declare.
3784 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3786 (native_encode_initializer): New function, moved from dwarf2out.c.
3787 Adjust to native_encode_expr compatible arguments, including dry-run
3788 and partial extraction modes. Don't handle STRING_CST.
3789 (can_native_interpret_type_p): No longer static.
3790 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3791 offset / BITS_PER_UNIT fits into int and don't call it if
3792 can_native_interpret_type_p fails. If suboff is NULL and for
3793 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3794 native_encode_initializer.
3795 (fold_const_aggregate_ref_1): Formatting fix.
3796 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3797 (tree_add_const_value_attribute): Adjust caller.
3799 PR tree-optimization/90838
3800 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3801 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3802 CTZ_DEFINED_VALUE_AT_ZERO.
3804 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3807 * lra-constraints.c (match_reload): Permit input operands have the
3808 same mode as output while other input operands have a different
3811 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3813 PR tree-optimization/90838
3814 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3815 (check_ctz_string): Likewise.
3816 (optimize_count_trailing_zeroes): Likewise.
3817 (simplify_count_trailing_zeroes): Likewise.
3818 (pass_forwprop::execute): Try ctz simplification.
3819 * match.pd: Add matching for ctz idioms.
3821 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3823 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3825 (aarch64_invalid_unary_op): New function for target hook.
3826 (aarch64_invalid_binary_op): New function for target hook.
3828 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3830 * config.gcc: Add arm_bf16.h.
3831 * config/aarch64/aarch64-builtins.c
3832 (aarch64_simd_builtin_std_type): Add BFmode.
3833 (aarch64_init_simd_builtin_types): Define element types for vector
3835 (aarch64_init_bf16_types): New function.
3836 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3837 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3839 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3840 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3842 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3843 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3844 * config/aarch64/aarch64.c
3845 (aarch64_classify_vector_mode): Add support for BF types.
3846 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3847 (aarch64_vq_mode): Add support for BF types.
3848 (aarch64_simd_container_mode): Add support for BF types.
3849 (aarch64_mangle_type): Add support for BF scalar type.
3850 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3851 * config/aarch64/arm_bf16.h: New file.
3852 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3853 * config/aarch64/iterators.md: Add BF types to mode attributes.
3854 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3856 2020-01-10 Jason Merrill <jason@redhat.com>
3858 PR c++/93173 - incorrect tree sharing.
3859 * gimplify.c (copy_if_shared): No longer static.
3860 * gimplify.h: Declare it.
3862 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3864 * doc/invoke.texi (-msve-vector-bits=): Document that
3865 -msve-vector-bits=128 now generates VL-specific code for
3866 little-endian targets.
3867 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3868 build_vector_type_for_mode to construct the data vector types.
3869 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3870 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3871 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3872 for 128-bit vectors.
3874 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3876 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3879 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3881 * config/aarch64/aarch64-builtins.c
3882 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3883 rather than checking the number of elements and the element mode.
3885 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3887 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3888 get_related_vectype_for_scalar_type rather than build_vector_type
3889 to create the index type for a conditional reduction.
3891 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3893 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3894 for any type of gather or scatter, including strided accesses.
3896 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3898 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3901 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3903 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3905 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3906 parameter and its use to reset DR_OFFSET's.
3907 (vect_transform_loop): Remove orig_drs_init argument.
3908 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3909 member of dr_vec_info rather than the offset of the associated
3910 data_reference's innermost_loop_behavior.
3911 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3912 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3913 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3914 get_dr_vinfo_offset.
3915 (vectorizable_store): Likewise.
3916 (vectorizable_load): Likewise.
3918 2020-01-10 Richard Biener <rguenther@suse.de>
3920 * gimple-ssa-store-merging
3921 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3923 2020-01-10 Martin Liska <mliska@suse.cz>
3926 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3927 encapsulation that was there before r280040.
3929 2020-01-10 Richard Biener <rguenther@suse.de>
3932 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3933 sequences to avoid walking them again for secondary opportunities.
3934 (pass_lower_eh_dispatch::execute): Instead actually insert
3937 2020-01-10 Richard Biener <rguenther@suse.de>
3940 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3941 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3942 avoid quadraticness.
3944 2020-01-10 Martin Jambor <mjambor@suse.cz>
3946 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3947 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3948 to get param_ipa_sra_max_replacements.
3949 (param_splitting_across_edge): Pass the caller to
3950 pull_accesses_from_callee.
3952 2020-01-10 Martin Jambor <mjambor@suse.cz>
3954 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3955 * ipa-cp.c (max_new_size): Removed.
3956 (orig_overall_size): New variable.
3957 (get_max_overall_size): New function.
3958 (estimate_local_effects): Use it. Adjust dump.
3959 (decide_about_value): Likewise.
3960 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3961 orig_overall_size. Adjust dump.
3962 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3964 2020-01-10 Martin Jambor <mjambor@suse.cz>
3966 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3967 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3968 instead of param_ipa_max_agg_items.
3969 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3970 optimization info for the callee.
3972 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3974 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3975 markers if debug_inline_points is false.
3977 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3979 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3981 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3982 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3983 aarch64-sve-builtins-sve2.h.
3984 (aarch64-sve-builtins-sve2.o): New rule.
3985 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3986 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3987 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3988 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3989 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3990 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3992 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3993 instructions that are handled here.
3994 (@cond_asrd<mode>): Generalize to...
3995 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3996 (*cond_asrd<mode>_2): Generalize to...
3997 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3998 (*cond_asrd<mode>_z): Generalize to...
3999 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4000 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4001 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4002 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4003 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4005 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4006 (@aarch64_scatter_stnt<mode>): Likewise.
4007 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4008 (@aarch64_mul_lane_<mode>): Likewise.
4009 (@aarch64_sve_suqadd<mode>_const): Likewise.
4010 (*<sur>h<addsub><mode>): Generalize to...
4011 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4013 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4014 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4015 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4016 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4017 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4018 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4019 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4020 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4021 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4022 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4023 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4024 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4025 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4026 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4027 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4028 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4029 (@aarch64_sve2_xar<mode>): Likewise.
4030 (@aarch64_sve2_bcax<mode>): Likewise.
4031 (*aarch64_sve2_eor3<mode>): Rename to...
4032 (@aarch64_sve2_eor3<mode>): ...this.
4033 (@aarch64_sve2_bsl<mode>): New expander.
4034 (@aarch64_sve2_nbsl<mode>): Likewise.
4035 (@aarch64_sve2_bsl1n<mode>): Likewise.
4036 (@aarch64_sve2_bsl2n<mode>): Likewise.
4037 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4038 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4039 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4040 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4041 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4042 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4043 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4044 (<su>mull<bt><Vwide>): Generalize to...
4045 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4047 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4048 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4049 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4050 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4051 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4052 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4053 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4054 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4055 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4056 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4057 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4058 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4059 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4060 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4061 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4062 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4063 (<SHRNB:r>shrnb<mode>): Generalize to...
4064 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4066 (<SHRNT:r>shrnt<mode>): Generalize to...
4067 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4069 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4070 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4071 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4072 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4073 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4074 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4075 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4076 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4077 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4078 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4079 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4080 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4081 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4082 (@aarch64_sve2_cvtnt<mode>): Likewise.
4083 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4084 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4085 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4086 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4087 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4088 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4089 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4090 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4091 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4092 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4093 (@aarch64_sve2_pmul<mode>): Likewise.
4094 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4095 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4096 (@aarch64_sve2_tbl2<mode>): Likewise.
4097 (@aarch64_sve2_tbx<mode>): Likewise.
4098 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4099 (@aarch64_sve2_histcnt<mode>): Likewise.
4100 (@aarch64_sve2_histseg<mode>): Likewise.
4101 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4102 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4103 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4104 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4105 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4106 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4107 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4108 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4109 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4110 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4111 (SVE2_PMULL_PAIR_I): New mode iterators.
4112 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4113 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4114 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4115 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4116 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4117 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4118 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4119 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4120 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4121 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4122 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4123 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4124 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4125 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4126 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4127 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4128 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4129 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4130 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4131 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4132 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4133 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4134 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4135 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4136 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4137 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4138 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4139 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4140 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4141 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4142 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4144 (VNARROW, Ventype): New mode attributes.
4145 (Vewtype): Handle VNx2DI. Fix typo in comment.
4146 (VDOUBLE): New mode attribute.
4147 (sve_lane_con): Handle VNx8HI.
4148 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4149 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4150 (sve_int_op, sve_int_op_rev): Handle the above codes.
4151 (sve_pred_int_rhs2_operand): Likewise.
4152 (MULLBT, SHRNB, SHRNT): Delete.
4153 (SVE_INT_SHIFT_IMM): New int iterator.
4154 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4155 and UNSPEC_WHILEHS for TARGET_SVE2.
4156 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4157 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4158 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4159 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4160 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4161 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4162 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4163 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4164 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4165 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4166 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4167 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4168 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4169 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4170 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4171 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4172 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4173 (optab): Handle the new unspecs.
4174 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4176 (lr): Handle the new unspecs.
4178 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4179 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4180 (sve_int_qsub_op): New int attributes.
4181 (sve_fp_op, rot): Handle the new unspecs.
4182 * config/aarch64/aarch64-sve-builtins.h
4183 (function_resolver::require_matching_pointer_type): Declare.
4184 (function_resolver::resolve_unary): Add an optional boolean argument.
4185 (function_resolver::finish_opt_n_resolution): Add an optional
4186 type_suffix_index argument.
4187 (gimple_folder::redirect_call): Declare.
4188 (gimple_expander::prepare_gather_address_operands): Add an optional
4190 * config/aarch64/aarch64-sve-builtins.cc: Include
4191 aarch64-sve-builtins-sve2.h.
4192 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4193 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4194 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4195 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4196 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4197 (TYPES_s_unsigned): Likewise.
4198 (TYPES_s_integer): Use TYPES_s_unsigned.
4199 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4200 (TYPES_sd_integer): Use them.
4201 (TYPES_d_unsigned): New macro.
4202 (TYPES_d_integer): Use it.
4203 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4204 (TYPES_cvt_narrow): Likewise.
4205 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4206 (preds_mx): New variable.
4207 (function_builder::add_overloaded_function): Allow the new feature
4208 set to be more restrictive than the original one.
4209 (function_resolver::infer_pointer_type): Remove qualifiers from
4210 the pointer type before printing it.
4211 (function_resolver::require_matching_pointer_type): New function.
4212 (function_resolver::resolve_sv_displacement): Handle functions
4213 that don't support 32-bit vector indices or svint32_t vector offsets.
4214 (function_resolver::finish_opt_n_resolution): Take the inferred type
4215 as a separate argument.
4216 (function_resolver::resolve_unary): Optionally treat all forms in
4217 the same way as normal merging functions.
4218 (gimple_folder::redirect_call): New function.
4219 (function_expander::prepare_gather_address_operands): Add an argument
4220 that says whether scaled forms are available. If they aren't,
4221 handle scaling of vector indices and don't add the extension and
4223 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4224 fall back to using cond_* instead.
4225 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4226 Split out the member variables into...
4227 (rtx_code_function_base): ...this new base class.
4228 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4229 (unspec_based_function): Split out the member variables into...
4230 (unspec_based_function_base): ...this new base class.
4231 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4232 (unspec_based_function_exact_insn): New class.
4233 (unspec_based_add_function, unspec_based_add_lane_function)
4234 (unspec_based_lane_function, unspec_based_pred_function)
4235 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4236 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4237 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4239 (unspec_based_fused_function): New class.
4240 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4241 (unspec_based_fused_lane_function): New class.
4242 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4244 (CODE_FOR_MODE1): New macro.
4245 (fixed_insn_function): New class.
4246 (while_comparison): Likewise.
4247 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4248 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4249 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4250 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4251 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4252 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4253 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4254 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4255 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4256 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4257 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4258 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4259 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4260 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4261 Also add an initial argument for unary_convert_narrowt, regardless
4262 of the predication type.
4263 (build_32_64): Allow loads and stores to specify MODE_none.
4264 (build_sv_index64, build_sv_uint_offset): New functions.
4265 (long_type_suffix): New function.
4266 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4267 (binary_imm_long_base, load_gather_sv_base): Likewise.
4268 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4269 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4270 (unary_narrowb_base, unary_narrowt_base): Likewise.
4271 (binary_long_lane_def, binary_long_lane): New shape.
4272 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4273 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4274 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4275 (binary_to_uint_def, binary_to_uint): Likewise.
4276 (binary_wide_def, binary_wide): Likewise.
4277 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4278 (compare_def, compare): Likewise.
4279 (compare_ptr_def, compare_ptr): Likewise.
4280 (load_ext_gather_index_restricted_def,
4281 load_ext_gather_index_restricted): Likewise.
4282 (load_ext_gather_offset_restricted_def,
4283 load_ext_gather_offset_restricted): Likewise.
4284 (load_gather_sv_def): Inherit from load_gather_sv_base.
4285 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4286 (shift_left_imm_def, shift_left_imm): Likewise.
4287 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4288 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4289 (store_scatter_index_restricted_def,
4290 store_scatter_index_restricted): Likewise.
4291 (store_scatter_offset_restricted_def,
4292 store_scatter_offset_restricted): Likewise.
4293 (tbl_tuple_def, tbl_tuple): Likewise.
4294 (ternary_long_lane_def, ternary_long_lane): Likewise.
4295 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4296 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4297 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4298 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4299 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4300 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4301 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4302 (ternary_uint_def, ternary_uint): Likewise.
4303 (unary_convert): Fix typo in comment.
4304 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4305 (unary_long_def, unary_long): Likewise.
4306 (unary_narrowb_def, unary_narrowb): Likewise.
4307 (unary_narrowt_def, unary_narrowt): Likewise.
4308 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4309 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4310 (unary_to_int_def, unary_to_int): Likewise.
4311 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4312 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4313 (svasrd_impl): Delete.
4314 (svcadd_impl::expand): Handle integer operations too.
4315 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4316 new functions to derive the unspec numbers.
4317 (svmla_svmls_lane_impl): Replace with...
4318 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4319 integer operations too.
4320 (svwhile_impl): Rename to...
4321 (svwhilelx_impl): ...this and inherit from while_comparison.
4322 (svasrd): Use unspec_based_function.
4323 (svmla_lane): Use svmla_lane_impl.
4324 (svmls_lane): Use svmls_lane_impl.
4325 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4326 (svwhilele, svwhilelt): Use svwhilelx_impl.
4327 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4328 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4329 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4330 * config/aarch64/aarch64-sve-builtins.def: Include
4331 aarch64-sve-builtins-sve2.def.
4333 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4335 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4336 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4337 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4338 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4339 immediates as well as vector ones.
4340 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4341 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4342 (aarch64_sve_qsub_immediate): Update calls accordingly.
4344 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4346 * config/aarch64/aarch64-sve2.md: Add banner comments.
4347 (<su>mulh<r>s<mode>3): Move further up file.
4348 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4349 (*aarch64_sve2_sra<mode>): Move further down file.
4350 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4352 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4354 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4356 (while_optab_cmp): Handle them.
4357 * config/aarch64/aarch64-sve.md
4358 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4359 and add a "@" marker.
4360 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4361 instead of gen_aarch64_sve2_while_ptest.
4362 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4364 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4366 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4367 (UNSPEC_WHILELE): ...this.
4368 (UNSPEC_WHILE_LO): Rename to...
4369 (UNSPEC_WHILELO): ...this.
4370 (UNSPEC_WHILE_LS): Rename to...
4371 (UNSPEC_WHILELS): ...this.
4372 (UNSPEC_WHILE_LT): Rename to...
4373 (UNSPEC_WHILELT): ...this.
4374 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4375 (cmp_op, while_optab_cmp): Likewise.
4376 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4377 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4378 (svwhilelt): Likewise.
4380 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4382 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4383 (unary_to_uint): Define.
4384 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4385 (unary_count): Rename to...
4386 (unary_to_uint_def, unary_to_uint): ...this.
4387 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4389 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4391 * config/aarch64/aarch64-sve-builtins-functions.h
4392 (code_for_mode_function): New class.
4393 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4394 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4395 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4396 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4397 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4399 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4401 * config/aarch64/iterators.md (addsub): New code attribute.
4402 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4404 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4405 in the asm string and attributes. Fix indentation.
4406 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4408 (@aarch64_sve_<optab><mode>): ...this.
4409 * config/aarch64/aarch64-sve-builtins.h
4410 (function_expander::expand_signed_unpred_op): Delete.
4411 * config/aarch64/aarch64-sve-builtins.cc
4412 (function_expander::expand_signed_unpred_op): Likewise.
4413 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4414 try using code_for_aarch64_sve instead.
4415 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4416 (svqsub_impl): Likewise.
4417 (svqadd, svqsub): Use rtx_code_function instead.
4419 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4421 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4422 (HADDSUB, sur, addsub): Remove them.
4424 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4426 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4427 functions the same way, rather than singling out those that
4428 aren't mapped directly to optabs.
4430 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4432 * target.def (compatible_vector_types_p): New target hook.
4433 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4434 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4435 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4436 * doc/tm.texi: Regenerate.
4437 * gimple-expr.c: Include target.h.
4438 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4439 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4441 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4442 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4443 Use the original predicate if it already has a suitable type.
4445 2020-01-09 Martin Jambor <mjambor@suse.cz>
4447 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4448 resolve_speculation and redirect_call_stmt_to_callee static. Change
4449 return type of set_call_stmt to cgraph_edge *.
4450 * auto-profile.c (afdo_indirect_call): Adjust call to
4451 redirect_call_stmt_to_callee.
4452 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4453 make the this pointer explicit, adjust self-recursive calls and the
4454 call top make_direct. Return the resulting edge.
4455 (cgraph_edge::remove): Make this pointer explicit.
4456 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4457 (cgraph_edge::make_direct): Likewise, adjust call to
4458 resolve_speculation.
4459 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4460 call to set_call_stmt.
4461 (cgraph_update_edges_for_call_stmt_node): Update call to
4462 set_call_stmt and remove.
4463 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4464 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4465 (cgraph_node::create_edge_including_clones): Moved "first" definition
4466 of edge to the block where it was used. Adjusted calls to
4468 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4469 cgraph_edge::remove.
4470 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4471 make_direct and redirect_call_stmt_to_callee.
4472 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4473 resolve_speculation and make_direct.
4474 * ipa-inline-transform.c (inline_transform): Adjust call to
4475 redirect_call_stmt_to_callee.
4476 (check_speculations_1):: Adjust call to resolve_speculation.
4477 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4478 resolve-speculation.
4479 (inline_small_functions): Adjust call to resolve_speculation.
4480 (ipa_inline): Likewise.
4481 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4483 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4484 safe with regards to edge removal, adjust calls to
4485 redirect_call_stmt_to_callee.
4486 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4487 and redirect_call_stmt_to_callee.
4488 * multiple_target.c (create_dispatcher_calls): Adjust call to
4489 redirect_call_stmt_to_callee
4490 (redirect_to_specific_clone): Likewise.
4491 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4492 Adjust calls to cgraph_edge::remove.
4493 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4494 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4495 (expand_call_inline): Adjust call to cgraph_edge::remove.
4497 2020-01-09 Martin Liska <mliska@suse.cz>
4499 * params.opt: Set Optimization for
4500 param_max_speculative_devirt_maydefs.
4502 2020-01-09 Martin Sebor <msebor@redhat.com>
4506 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4508 2020-01-09 Martin Liska <mliska@suse.cz>
4510 * auto-profile.c (auto_profile): Use opt_for_fn
4512 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4513 (propagate_vals_across_arith_jfunc): Likewise.
4514 (hint_time_bonus): Likewise.
4515 (incorporate_penalties): Likewise.
4516 (good_cloning_opportunity_p): Likewise.
4517 (perform_estimation_of_a_value): Likewise.
4518 (estimate_local_effects): Likewise.
4519 (ipcp_propagate_stage): Likewise.
4520 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4521 (set_switch_stmt_execution_predicate): Likewise.
4522 (analyze_function_body): Likewise.
4523 * ipa-inline-analysis.c (offline_size): Likewise.
4524 * ipa-inline.c (early_inliner): Likewise.
4525 * ipa-prop.c (ipa_analyze_node): Likewise.
4526 (ipcp_transform_function): Likewise.
4527 * ipa-sra.c (process_scan_results): Likewise.
4528 (ipa_sra_summarize_function): Likewise.
4529 * params.opt: Rename ipcp-unit-growth to
4530 ipa-cp-unit-growth. Add Optimization for various
4531 IPA-related parameters.
4533 2020-01-09 Richard Biener <rguenther@suse.de>
4536 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4538 2020-01-09 Richard Biener <rguenther@suse.de>
4540 PR tree-optimization/93040
4541 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4543 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4545 * common/config/avr/avr-common.c (avr_option_optimization_table)
4546 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4548 2020-01-09 Martin Liska <mliska@suse.cz>
4550 * cgraphclones.c (symbol_table::materialize_all_clones):
4551 Use cgraph_node::dump_name.
4553 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4556 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4557 output_operand_lossage instead of gcc_unreachable.
4558 * doc/md.texi (riscv f constraint): Fix typo.
4561 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4562 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4563 CONST_SCALAR_INT_P instead of CONST_INT_P.
4564 (*subv<mode>4_1): Rename to ...
4565 (subv<mode>4_1): ... this.
4566 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4567 define_insn_and_split patterns.
4568 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4571 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4573 * vec.c (class selftest::count_dtor): New class.
4574 (selftest::test_auto_delete_vec): New test.
4575 (selftest::vec_c_tests): Call it.
4576 * vec.h (class auto_delete_vec): New class template.
4577 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4579 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4581 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4583 2020-01-08 Jim Wilson <jimw@sifive.com>
4585 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4586 use of TLS_MODEL_LOCAL_EXEC when not pic.
4588 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4590 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4593 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4596 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4597 *stack_protect_set_3 peephole2): Also check that the second
4598 insns source is general_operand.
4601 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4602 predicate for output operand instead of register_operand.
4603 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4604 memory destination and non-memory operands[2].
4606 2020-01-08 Martin Liska <mliska@suse.cz>
4608 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4609 ::dump_asm_name instead of (::name or ::asm_name).
4610 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4611 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4612 (analyze_functions): Likewise.
4613 (expand_all_functions): Likewise.
4614 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4615 (propagate_bits_across_jump_function): Likewise.
4616 (dump_profile_updates): Likewise.
4617 (ipcp_store_bits_results): Likewise.
4618 (ipcp_store_vr_results): Likewise.
4619 * ipa-devirt.c (dump_targets): Likewise.
4620 * ipa-fnsummary.c (analyze_function_body): Likewise.
4621 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4622 (process_hsa_functions): Likewise.
4623 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4624 (set_alias_uids): Likewise.
4625 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4626 * ipa-inline.c (recursive_inlining): Likewise.
4627 (inline_to_all_callers_1): Likewise.
4628 (ipa_inline): Likewise.
4629 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4630 (ipa_propagate_frequency): Likewise.
4631 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4632 (remove_described_reference): Likewise.
4633 * ipa-pure-const.c (worse_state): Likewise.
4634 (check_retval_uses): Likewise.
4635 (analyze_function): Likewise.
4636 (propagate_pure_const): Likewise.
4637 (propagate_nothrow): Likewise.
4638 (dump_malloc_lattice): Likewise.
4639 (propagate_malloc): Likewise.
4640 (pass_local_pure_const::execute): Likewise.
4641 * ipa-visibility.c (optimize_weakref): Likewise.
4642 (function_and_variable_visibility): Likewise.
4643 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4644 (ipa_discover_variable_flags): Likewise.
4645 * lto-streamer-out.c (output_function): Likewise.
4646 (output_constructor): Likewise.
4647 * tree-inline.c (copy_bb): Likewise.
4648 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4649 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4651 2020-01-08 Richard Biener <rguenther@suse.de>
4654 * tree-eh.c (sink_clobbers): Update virtual operands for
4655 the first and last stmt only. Add a dry-run capability.
4656 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4657 after CFG manipulations and in RPO order to catch all
4658 secondary opportunities reliably.
4660 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4663 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4665 2019-01-08 Richard Biener <rguenther@suse.de>
4668 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4669 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4670 virtual operand, also updating SSA use.
4671 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4672 Update stmt after resetting virtual operand.
4673 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4674 * gimple-iterator.c (gsi_remove): When not removing the stmt
4675 permanently do not delink immediate uses or mark the stmt modified.
4677 2020-01-08 Martin Liska <mliska@suse.cz>
4679 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4680 (ipa_call_context::estimate_size_and_time): Likewise.
4681 (inline_analyze_function): Likewise.
4683 2020-01-08 Martin Liska <mliska@suse.cz>
4685 * cgraph.c (cgraph_node::dump): Use systematically
4688 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4690 Add -nodevicespecs option for avr.
4693 * config/avr/avr.opt (-nodevicespecs): New driver option.
4694 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4695 "-specs=device-specs/..." if that option is not set.
4696 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4698 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4700 Implement 64-bit double functions for avr.
4703 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4704 --with-double-comparison.
4705 * doc/install.texi: Document them.
4706 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4707 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4708 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4709 * doc/invoke.texi (AVR Built-in Macros): Document them.
4710 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4711 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4712 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4714 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4717 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4718 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4719 when only building rm-profile multilibs.
4721 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4724 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4725 lattice for a value to check.
4726 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4727 finite propagation in self-recursive scc.
4729 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4731 * ipa-inline.c (caller_growth_limits): Restore the AND.
4733 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4735 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4736 (VEC_ALLREG_ALT): New iterator.
4737 (VEC_ALLREG_INT_MODE): New iterator.
4738 (VCMP_MODE): New iterator.
4739 (VCMP_MODE_INT): New iterator.
4740 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4741 (vec_cmp<u>v64qidi): New define_expand.
4742 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4743 (vec_cmpu<mode>di_exec): New define_expand.
4744 (vec_cmp<u>v64qidi_exec): New define_expand.
4745 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4746 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4747 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4748 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4749 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4750 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4751 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4752 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4753 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4754 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4756 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4757 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4759 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4761 * config/gcn/constraints.md (DA): Update description and match.
4763 (Db): New constraint.
4764 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4766 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4767 Implement 'Db' mixed immediate type.
4768 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4769 (addcv64si3_dup<exec_vcc>): Delete.
4770 (subcv64si3<exec_vcc>): Rework constraints.
4771 (addv64di3): Rework constraints.
4772 (addv64di3_exec): Rework constraints.
4773 (subv64di3): Rework constraints.
4774 (addv64di3_dup): Delete.
4775 (addv64di3_dup_exec): Delete.
4776 (addv64di3_zext): Rework constraints.
4777 (addv64di3_zext_exec): Rework constraints.
4778 (addv64di3_zext_dup): Rework constraints.
4779 (addv64di3_zext_dup_exec): Rework constraints.
4780 (addv64di3_zext_dup2): Rework constraints.
4781 (addv64di3_zext_dup2_exec): Rework constraints.
4782 (addv64di3_sext_dup2): Rework constraints.
4783 (addv64di3_sext_dup2_exec): Rework constraints.
4785 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4787 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4788 existing target checks.
4790 2020-01-07 Richard Biener <rguenther@suse.de>
4792 * doc/install.texi: Bump minimal supported MPC version.
4794 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4796 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4797 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4798 * langhooks.c: Include stor-layout.h.
4799 (lhd_simulate_enum_decl): New function.
4800 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4801 handle_arm_sve_h for the LTO frontend.
4802 (register_vector_type): Cope with null returns from pushdecl.
4804 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4806 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4807 (aarch64_sve::nvectors_if_data_type): Replace with...
4808 (aarch64_sve::builtin_type_p): ...this.
4809 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4810 (find_vector_type): Delete.
4811 (add_sve_type_attribute): New function.
4812 (lookup_sve_type_attribute): Likewise.
4813 (register_builtin_types): Add an "SVE type" attribute to each type.
4814 (register_tuple_type): Likewise.
4815 (svbool_type_p, nvectors_if_data_type): Delete.
4816 (mangle_builtin_type): Use lookup_sve_type_attribute.
4817 (builtin_type_p): Likewise. Add an overload that returns the
4818 number of constituent vector and predicate registers.
4819 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4820 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4821 instead of aarch64_sve_argument_p.
4822 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4823 (aarch64_pass_by_reference): Likewise.
4824 (aarch64_function_value_1): Likewise.
4825 (aarch64_return_in_memory): Likewise.
4826 (aarch64_layout_arg): Likewise.
4828 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4830 PR tree-optimization/93156
4831 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4832 least significant bit is always clear.
4834 PR tree-optimization/93118
4835 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4836 simplifier with two intermediate conversions.
4838 2020-01-07 Martin Liska <mliska@suse.cz>
4840 * params.opt: Add Optimization for various parameters.
4842 2020-01-07 Martin Liska <mliska@suse.cz>
4845 * doc/extend.texi: Explain cloning for target_clone
4848 2020-01-07 Martin Liska <mliska@suse.cz>
4850 PR tree-optimization/92860
4851 * common.opt: Make in Optimization option
4852 as it is affected by -O0, which is an Optimization
4854 * tree-inline.c (tree_inlinable_function_p):
4855 Use opt_for_fn for warn_inline.
4856 (expand_call_inline): Likewise.
4858 2020-01-07 Martin Liska <mliska@suse.cz>
4860 PR tree-optimization/92860
4861 * common.opt: Make flag_ree as optimization
4864 2020-01-07 Martin Liska <mliska@suse.cz>
4866 PR optimization/92860
4867 * params.opt: Mark param_min_crossjump_insns with Optimization
4870 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4872 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4873 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4875 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4877 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4878 helper function to return the valid addressing formats for a given
4879 hard register and mode.
4880 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4882 * config/rs6000/constraints.md (Q constraint): Update
4884 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4887 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4888 Use 'Q' for doing vector extract from memory.
4889 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4891 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4892 doing vector extract from memory.
4893 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4894 extract from memory.
4896 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4897 for the offset being 34-bits when -mcpu=future is used.
4899 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4901 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4902 instead of cmpib_comparison_operator in cmpib patterns.
4903 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4904 of cmpib_comparison_operator. Revise comment.
4906 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4908 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4909 in an IFN_DIV_POW2 node to be equal.
4911 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4913 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4914 (vect_check_scalar_mask): ...this.
4915 (vectorizable_store, vectorizable_load): Update call accordingly.
4916 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4917 argument in calls to conditional internal functions.
4919 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4921 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4922 '0' matching inputs.
4923 (subv64di3_exec): Likewise.
4925 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4927 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4928 * doc/md.texi (movstr): Likewise.
4930 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4932 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4935 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4937 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4939 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4940 to a temporary file and use move-if-change to update the real
4941 file where necessary.
4943 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4945 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4946 rather than Upa for CPY /M.
4948 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4950 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4953 2020-01-06 Martin Liska <mliska@suse.cz>
4955 PR tree-optimization/92860
4956 * params.opt: Mark param_max_combine_insns with Optimization
4959 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4962 * config/i386/i386.md (SWIDWI): New mode iterator.
4963 (DWI, dwi): Add TImode variants.
4964 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4965 <general_hilo_operand> instead of <general_operand>. Use
4966 CONST_SCALAR_INT_P instead of CONST_INT_P.
4967 (*addv<mode>4_1): Rename to ...
4968 (addv<mode>4_1): ... this.
4969 (QWI): New mode attribute.
4970 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4971 define_insn_and_split patterns.
4972 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4974 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4975 <general_hilo_operand> instead of <general_operand>.
4976 (*addcarry<mode>_1): New define_insn.
4977 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4979 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4981 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4982 Use "call" instead of "set".
4984 2020-01-03 Martin Jambor <mjambor@suse.cz>
4987 * ipa-cp.c (print_all_lattices): Skip functions without info.
4989 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4992 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4993 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4994 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4995 for 'e' simd clones.
4998 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5000 (mprefer-vector-width=): Add Save.
5001 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5002 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5003 (ix86_debug_options, ix86_function_specific_print): Adjust
5004 ix86_target_string callers.
5005 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5006 (ix86_valid_target_attribute_tree): Likewise.
5007 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5008 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5009 ix86_target_string caller.
5012 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5013 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5014 instead of gen_int_shift_amount + convert_modes.
5016 PR rtl-optimization/93088
5017 * loop-iv.c (find_single_def_src): Punt after looking through
5018 128 reg copies for regs with single definitions. Move definitions
5021 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5023 * config/arm/arm-c.c (arm_cpu_builtins): Define
5024 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5025 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5026 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5027 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5028 * config/arm/arm-tables.opt: Regenerated.
5029 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5030 arm_arch_i8mm and arm_arch_bf16 when enabled.
5031 * config/arm/arm.h (TARGET_I8MM): New macro.
5032 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5033 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5034 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5035 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5036 (v8_6_a_simd_variants): New.
5037 (v8_*_a_simd_variants): Add i8mm and bf16.
5038 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5040 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5043 * predict.c (compute_function_frequency): Don't call
5044 warn_function_cold on functions that already have cold attribute.
5046 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5049 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5050 COMDAT group function labels in .data.rel.ro.local section.
5051 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5054 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5055 comparison_operator in B and S integer comparisons. Likewise, use
5056 ordered_comparison_operator instead of cmpib_comparison_operator in
5058 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5060 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5062 Update copyright years.
5064 * gcc.c (process_command): Update copyright notice dates.
5065 * gcov-dump.c (print_version): Ditto.
5066 * gcov.c (print_version): Ditto.
5067 * gcov-tool.c (print_version): Ditto.
5068 * gengtype.c (create_file): Ditto.
5069 * doc/cpp.texi: Bump @copying's copyright year.
5070 * doc/cppinternals.texi: Ditto.
5071 * doc/gcc.texi: Ditto.
5072 * doc/gccint.texi: Ditto.
5073 * doc/gcov.texi: Ditto.
5074 * doc/install.texi: Ditto.
5075 * doc/invoke.texi: Ditto.
5077 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5079 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5082 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5084 PR tree-optimization/93098
5085 * match.pd (popcount): For shift amounts, use integer_onep
5086 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5087 tests. Make sure that precision is power of two larger than or equal
5088 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5089 instead of ULL suffixed constants. Formatting fixes.
5091 Copyright (C) 2020 Free Software Foundation, Inc.
5093 Copying and distribution of this file, with or without modification,
5094 are permitted in any medium without royalty provided the copyright
5095 notice and this notice are preserved.